EXPERIMENTAL STUDIES AND THEORETICAL MODELING OF CONCRETE SUBJECTED TO HIGH TEMPERATURES by JAESUNG LEE B.A.

, Architectural Engineering, Hannam University, Korea, 1997 M.A., Architectural Engineering, Hanyang University, Korea, 1999 M.A., Civil Engineering, University of Colorado at Boulder, USA, 2002

A thesis submitted to the Faculty of the Graduate school of the University of Colorado in partial fulfillment of the requirement for the degree of Doctor of Philosophy Department of Civil, Environmental, and Architectural Engineering 2006

iii Lee, Jaesung (Ph.D., Civil, Environmental, and Architectural Engineering) EXPERIMENTAL STUDIES AND THEORETICAL MODELING OF

CONCRETE SUBJECTED TO HIGH TEMPERATURES Thesis directed by Professor Yunping Xi

Understanding the properties of concrete under high temperature is essential to enhance the fire resistance of reinforced concrete structures (RCS) and to provide accurate information for fire design of RCS. Extensive studies on this important topic were performed previously. However, the properties of concrete under high temperature have not been fully understood. Even if there are numerous experimental and theoretical results available in the literature, contradictions among observations exist and need to be reconciled. The studies performed in this thesis can be largely classified as four topics. Among the four topics, two topics are experimental studies and the other two topics are theoretical models. The first experimental study is to find the effects of temperature and moisture on deformation of concrete. Due to the dependency of the moisture on temperature, it is not easy to distinguish experimentally the effects of temperature and moisture on the deformation of concrete. Usually, the thermal strain of concrete, which is called Conventional Thermal Strain (CTS) in this study, is obtained by measuring the displacement change without moisture control. In this study, CTS, the strain caused by temperature increase under constant humidity (Pure Thermal Strain: PTS), and the strain caused by moisture change under constant temperature (Pure Hygro Strain:

iv PHS) are measured continually over time. From the data analysis based on the measured strains, the thermo-hygro coupling effect in the temperature is obtained. Previous experimental studies on concrete under high temperatures have mainly concentrated on the strength reduction of concrete, even though the loss of durability of concrete can severely reduce the remaining service life of the structure. In the second experimental study, the strength, stiffness, and durability performance of concrete subjected to various heating and cooling scenarios are investigated. Variations of concrete under high temperatures result mainly from two mechanisms. One is the variation of material properties of the constituent phases under high temperatures, and the other is the transformation of constituent phases under different temperatures. Therefore, the properties of concrete under high temperatures must be studied from both mechanical and chemical points of view. The model for the thermal degradation of elastic modulus of concrete is obtained by composite damage mechanics. At the level of cement paste, the variations of volume fractions of the constituents are based on phase transformations taking place under different temperature ranges. At the levels of mortar and concrete, the degradations of various sand and gravel are based on available test data. The model for the thermal strain of cement paste and concrete, considering micro-structural changes under elevated temperatures ranging from 20C to 800C, is proposed. The model is a combination of a multi-scale stoichiometric model and a multi-scale composite model. The model can consider characteristics of various aggregates in the calculation of thermal expansion for concrete.

DEDICATION

To my lovely wife, my daughter, my son Myungsub Kim, Jihyun Lee, and Kisuk Lee my parents Sangyoung Lee and Inhee Kim my parents-in-law Sangyeal Kim and Soonja Bok my sisters and brothers Janghyun Baek, Kyungsook Lee, Ohsung Kwon, Nammi Lee, Bangsub Lee, and Myunghwa Park

vi

ACKNOWLEDGEMENTS

I appreciate the assistance, guidance, and inspiration of many people without whom this work would not have been possible. Professor Yunping Xi, my thesis advisor, provided guidance and support for this thesis and for my professional development. I am very grateful to him. I would also like to thank Professors Kaspar J. Willam, Ross B. Corotis, Kevin L. Rens, and Mettupalayam V. Sivaselvan for their helpful comments regarding this dissertation. I would like to thank my fellow students at the University of Colorado for their support, encouragement, and friendship. I am thankful to Keunkwang Lee and Raeyoung Jung for their sincere encouragement and advice through my student life. I am thankful to Jaemo Jung, Sunyoung Kim and Junsub Kim for their love. I am very grateful to my parents (Sangyoung Lee and Inhee Kim), parents-in-law (Sangyeal Kim and Soonja Bok), brothers-in-law (Janghyun Baek and Ohsung Kwon), sisters (Kyungsook Lee and Nammi Lee), brother (Bangsub Lee), sister-in-law (Myunghaw Park) for their faith, life-long support, and love. The support of the National Science Foundation through grants CMS-0409747 is also gratefully acknowledged. Finally, I would like to extend my deepest thanks to my wife (Myungsub Kim), my daughter (Jihyun Lee), and my son (Kisuk Lee) for their patience, encouragement and love.

vii

CONTENTS

CHAPTER 1. 1.1. 1.2. 1.3. 2. 2.1. INTRODUCTION ......................................................................................... 1 Background .................................................................................................... 1 Thesis objectives............................................................................................ 2 Thesis organization ........................................................................................ 3 LITERATURE REVIEW .............................................................................. 5 Microstructure variations of concrete due to high temperature ..................... 5 Microstructure of hydrated cement paste................................................. 5 Microstructure changes of cement paste due to high temperature......... 10 Porosity and pore size distribution according to temperature................ 13 Heat deformation of cement paste due to high temperature .................. 15 SEM (Scanning Electron Microscopy) observation of concrete............ 22 SEM observation of concrete used in current research.......................... 25

2.1.1. 2.1.2. 2.1.3. 2.1.4. 2.1.5. 2.1.6. 2.2.

Pure concrete exposed to high temperature ................................................. 27 Two major problems when concrete is exposed to fire ......................... 27 Selection of concrete mix....................................................................... 28 Testing methods for mechanical properties of concrete ........................ 35 Concrete properties under high temperatures ........................................ 39

2.2.1. 2.2.2. 2.2.3. 2.2.4.

viii 2.2.5. 2.2.6. 2.2.7. 2.2.8. 2.3. The effects of sustained loading during heating .................................... 47 Combined thermal and mechanical loading histories ............................ 49 Decomposition of the total strain due to coupling effects ..................... 51 The effects of steel and polymer fibers at elevated temperatures.......... 54

Concrete Structures under High Temperatures............................................ 56 Factors affecting fire performance of concrete structures ..................... 58 Fire scenarios ......................................................................................... 60 Experimental studies on concrete beams ............................................... 63 Experimental studies on concrete columns............................................ 72 Experimental studies on concrete slabs ................................................. 79

2.3.1. 2.3.2. 2.3.3. 2.3.4. 2.3.5. 3.

AN EXPERIMENTAL STUDY FOR THERMO-HYGRO COUPLING

EFFECT OF CONCRETE AT ELEVATED TEMPERATURE................................ 87 3.1. 3.2. Introduction.................................................................................................. 87 Experimental details..................................................................................... 88 Material and mixture portion ................................................................. 88 Specimen preparation and test equipments............................................ 91 Thermal compensation function and verification for test method ......... 94

3.2.1. 3.2.2. 3.2.3. 3.3.

Test results and discussion........................................................................... 97 Conventional Thermal Strain (CTS) and Pure Hygro Strain (PHS) ...... 97 Pure Thermal Strain (PTS)................................................................... 108 Thermo-hygro coupling effect ............................................................. 111

3.3.1. 3.3.2. 3.3.3. 3.4.

Conclusions................................................................................................ 114

ix 4. STRENGTH AND DURABILITY OF CONCRETE SUBJECTED TO

VARIOUS HEATING AND COOLING TREATMENTS...................................... 116 4.1. 4.2. 4.3. Introduction................................................................................................ 116 Specimen preparation, heating equipment and test variables .................... 117 Temperature distribution and thermal diffusivity ...................................... 123 Test set-up............................................................................................ 123 Results and discussion ......................................................................... 124

4.3.1. 4.3.2. 4.4.

Water permeability..................................................................................... 128 Test set-up............................................................................................ 128 Results and discussion ......................................................................... 130

4.4.1. 4.4.2. 4.5.

Ultrasonic Pulse Velocity (UPV) and residual compression test............... 139 Test equipments and test set-up ........................................................... 139 Results and discussion for UPV (Ultrasonic pulse velocity) ............... 140 Results and discussion for residual compression test .......................... 143 Comparison between Ep and Ei........................................................... 150

4.5.1. 4.5.2. 4.5.3. 4.5.4. 4.6. 4.7. 4.8. 5.

Weight loss................................................................................................. 152 Cracks and color changes of the specimens............................................... 155 Conclusions................................................................................................ 157 A MULTISCALE CHEMO-MECHANICAL MODEL FOR MODULUS

OF ELASTICITY OF CONCRETE UNDER HIGH TEMPERATURES ............... 159 5.1. 5.2. 5.3. Introduction................................................................................................ 159 Hydration kinetics model........................................................................... 162 Initial volume fractions of constituent phases in concrete......................... 166

x 5.3.1. 5.3.2. 5.4. Volume fractions of the constituent phases at the cement paste level. 166 Volume fractions of the phases at the mortar and concrete levels....... 171

A multi-scale stoichiometric model for phase transformations ................. 172 Phase transformations at the cement paste level.................................. 172 Validation of current model at cement paste level............................... 177 Volume fractions of the phases according to temperature increase at the

5.4.1. 5.4.2. 5.4.3.

mortar and concrete level.................................................................................. 181 5.5. 5.6. 5.7. 5.8. 6. STRAIN Composite theories and damage theories................................................... 182 Thermal degradation of the modulus elasticity of concrete....................... 185 Comparison between present model and experimental results .................. 188 Conclusions................................................................................................ 192 A MULTISCALE CHEMO-MECHANICAL MODEL FOR THERMAL OF CEMENT PASTE AND CONCRETE UNDER HIGH

TEMPERATURES ................................................................................................... 195 6.1. 6.2. 6.3. 6.4. 6.5. 7. Introduction................................................................................................ 195 Model for shrinkage of cement paste and concrete ................................... 196 Material properties of constituents............................................................. 199 Comparison between model and experimental results............................... 206 Conclusions................................................................................................ 211 SUMMARY AND CONCLUSIONS ........................................................ 213

REFERNCES............................................................................................................ 220 APPENDIX: RELATIVE HUMIDITY IN UNSATURATED SOIL ...................... 237

xi

TABLES

TABLE Table. 2.1. Porosity and Pore Size Distribution (Piasta et al, 1984b)......................... 13 Table 2.2. Temperature regimes of chemical reactions of hydrate phase of hardened cement paste according to Schneider and Herbst (1989).................................... 14 Table 2.3. A summary for heat deformation of cement paste..................................... 21 Table 2.4. Factors influencing spalling of concrete (Khoury, 2000) .......................... 30 Table 2.5. Mix design (Dotreppe et al., 1996) ............................................................ 73 Table 3.1. Typical proportions of materials for normal weight concretes.................. 89 Table 4.1 Coefficients of water permeability for R15_D2 test series....................... 135 Table 4.2 Coefficients of water permeability for R2_D4 test series......................... 135 Table 4.3. Ultrasonic pulse velocity test result ......................................................... 141 Table 4.4. Unit weight measured from specimens used in present study ................. 142 Table 4.5. Elastic modulus ( E p ) from ultrasonic pulse velocity.............................. 142 Table 4.6. Ultimate strength results of residual compression test ........................... 147 Table 4.7. Relative ultimate strength ........................................................................ 148 Table 4.8. Initial tangent modulus from residual compression test .......................... 149 Table 4.9. Relative initial tangent modulus .............................................................. 149 Table 4.10. Effect of heating on weight loss (Initial relative humidity 90%) .......... 153

xii Table 4.11. Effect of heating on weight loss (RH 100%:Fully saturated)................ 153 Table 5.1. Coefficients of ai , bi , and ci .................................................................. 164 Table 5.2. Diffusion constant and coefficients in the diffusion model..................... 165 Table 5.3. Parameters for the determination of the volume fractions....................... 169 Table 5.4. Volume fractions of constituents at cement paste level (w/c=0.5 and 0.67) ........................................................................................................................... 170 Table 5.5. Density of different rock groups (Road research laboratory, 1959) ........ 172 Table 5.6. Volume fractions at mortar and concrete level (w/c=0.67) ..................... 172 Table 5.7. Processes of decomposition depending on the temperature regime ........ 174 Table 5.8. Theoretical formulas for volume fraction change of each phase (w/c=0.67) ........................................................................................................................... 176 Table 5.9. Mass fractions of clinker phases in the cement ....................................... 177 Table. 5.10. Porosity and pore size distribution [w/c=0.4, (Piasta et al, 1984)]....... 178 Table 5.11. Summary for the results shown in Fig. 5.4 and Fig. 5.5........................ 180 Table 5.12. Volume fraction of phases at mortar level (w/c=0.67) .......................... 181 Table 5.13. Volume fraction of phases at concrete level (w/c=0.67) ....................... 181 Table 5.14. Elastic properties of constituent phases ................................................. 188 Table 5.15. Elastic moduli of each phase used in the present model........................ 189 Table 5.16. Elastic moduli from residual compression test ...................................... 191 Table 5.17. Relative elastic modulus ........................................................................ 191 Table 6.1. Summary for thermal strain of phases in cement paste used in model.... 203 Table 6.2. Elastic modulus and Porosity of each phase............................................ 204 Table 6.3. Elastic properties of various aggregates (Jumijis, 1983) ......................... 205

xiii Table 6.4. Summary for thermal strain of limestone and sandstone......................... 206 Table 6.5. Thermal strain functions of cement paste from model (w/c=0.5)............ 208

xiv

FIGURES

FIGURE Figure 2.1. SEM image of tobermorite (C-S-H, Type I) .............................................. 7 Figure 2.2. SEM image of portlandite (CH) crystals .................................................... 7 Figure 2.3. SEM images of ettringite crystals............................................................... 8 Figure 2.4. Model of a well-hydrated Portland cement paste ....................................... 9 Figure 2.5. Ca (OH ) 2 contents observed from TG and X-ray (Piasta et al, 1984 b).. 11 Figure 2.6. SEM images of calcite crystals................................................................. 12 Figure 2.7. Thermograms of unhydrated clinker minerals and Portland cement: 1) C3S , 2) -C2 S , 3) C3 A , 4) C4 AF , 5) Portland cement ............................... 16 Figure 2.8. Thermograms of hydrated materials: 1) C3S H , 2) -C2 S H , 3) C3 AH , 4) C4 AFH , 5) Portland cement paste, 6) Ca (OH ) 2 , 7) Ettringite ...................... 17 Figure 2.9. 1) Ca (OH ) 2 , 2) cement gel [300 C exposure (1000), Piasta (1984a)] 19 Figure 2.10. Destroyed grains of unhydrated clinker as a result of different thermal deformations [500 C exposure (3000), Piasta (1984a)] .................................. 20 Figure 2.11. Morphologies of hydrates after 400 C exposure (Lin et al, 1996) ........ 22 Figure 2.12. Morphologies of hydrates after 500 C exposure (Lin et al, 1996) ........ 23 Figure 2.13. Crakes and honey combs after 900 C exposure (Lin et al, 1996) ......... 23 Figure 2.14. Sample under water cooling after 900 C exposure (Lin et al, 1996) .... 24

xv Figure 2.15. SEM images of the concrete used in current research............................ 26 Figure 2.16. Physicochemical process in Portland concrete during (Khoury, 2000) 27 Figure 2.17. Material choices for concrete under high temperature (Khoury, 2000) . 28 Figure 2.18. Mechanism of pore pressure spalling (Schneider and Horbst, 2002)..... 32 Figure 2.19. Gradients of temperature, pore pressure and moisture in a massive concrete section heated at the unsealed left-hand surface (Khoury, 2000)......... 32 Figure 2.20. Forces acting in heated concrete (Zhukov, 1975) .................................. 34 Figure 2.21. Mechanical property test setup (Phan, 2002) ......................................... 34 Figure 2.22. Test set up, Pore pressure specimen, and Specimen failure (Phan, 2002) ............................................................................................................................. 35 Figure 2.23. Schematic of temperature and loading histories for the three test methods ............................................................................................................................. 38 Figure 2.24. Hot isothermal test results obtained by Castillo and Durani (1990) ...... 40 Figure 2.25. Hot isothermal test results obtained by Abrams (1971) ......................... 42 Figure 2.26. Hot isothermal test results obtained by Morita et al. (1992) .................. 43 Figure 2.27. Hot isothermal test results obtained by Furumura et al. (1995) ............. 44 Figure 2.28. Hot isothermal test results obtained by Noumowe et al. (1996) ............ 46 Figure 2.29. The effects of loading and temperature during heating in uniaxial compression of unsealed concrete specimens (Khoury, 2002)........................... 48 Figure 2.30. Effect of temperature upon the residual (after cooling) compressive strength and elastic modulus of unsealed C70 HITECO Concrete -20 percent load: expressed as a percentage of strength prior to heating (Khoury, 2002)..... 49

xvi Figure 2.31. Measured total strain in concrete specimens heated to 400 C. Applied stress is 45 % of compressive strength at 20 C (Anderberg and Thelandersson, 1976) ................................................................................................................... 50 Figure 2.32. Strain of unsealed basalt concrete measured during heating at 1 C/min under three uniaxial compressive load levels (percent of strength prior to heating) excluding the initial elastic strain (Khoury, 2002)........................................... 52 Figure 2.33. Relative proportions of three load induced strains in uniaxial compression of concrete (Khoury, 2002)............................................................ 52 Figure 2.34. Failure mode with steel fibers (a) and without steel fibers (b) in HSC.. 55 Figure 2.35. The loading frame and the furnace (Fu-ping Cheng et al., 2004) .......... 55 Figure 2.36. High strength concrete blocks after two-hour hydrocarbon fire test:..... 56 Figure 2.37. A fire test of full-scale steel structure built in Cardington Laboratory (Grosshandler, 2002)........................................................................................... 57 Figure 2.38. Tie configuration for reinforced concrete column: (a) conventional tie configuration (b) modified tie configuration (Kodur, 1999) .............................. 60 Figure 2.39. Standard fire curves for (ISO 834 (or BS476) and ASTM E119) based on a typical building fire (Grosshandler, 2002) ....................................................... 61 Figure 2.40. Standard fire scenarios for buildings (ISO 834 or BS 476), offshore and petrochemical industries (hydrocarbon), and tunnels (RWS, RABT) (Khoury, 2000) ................................................................................................................... 61 Figure 2.41. Temperature vs. Time at different locations (Sanjayan and Stocks, 1991) ............................................................................................................................. 64 Figure 2.42. Weight loss vs. Time in the test (Sanjayan and Stocks, 1991).............. 64

xvii Figure 2.43. ISO Hydrocarbon Fire Curve, Furnace Temperature, and Time Period when Spalling was Observed (Hansen and Jensen, 1995).................................. 66 Figure 2.44. Specimen geometry and thermal cycle (Felicetti and Gambarova, 1999) ............................................................................................................................. 67 Figure 2.45. Strength of the three mix designs (Felicetti and Gambarova, 1999)...... 68 Figure 2.46. Fringe pattern and cracking of specimen at T=20 C ............................. 69 Figure 2.47. Cracking at failure by mechanical load after one cycle heating up to 500 C (Felicetti and Gambarova, 1999) ............................................................ 69 Figure 2.48. Load-displacement curves at bottom face of mid span and crack patterns (Felicetti and Gambarova, 1999) ........................................................................ 70 Figure 2.49. (a) Flexural model; (b) strut-and-tie model; (c) beam-end model........ 71 Figure 2.50. Comparison of failure models (Felicetti and Gambarova, 1999) ........... 71 Figure 2.51. Furnace used for tests on columns at the University of Ghent: (a) Overall view (b) Detail of top hinge and load cell (2000 kN) (c) Detail of bottom hinge (measurements in millimeters) (Dotreppe at el., 1996) ...................................... 74 Figure 2.52. Furnace used for tests on columns at the University of Lige: (1) Lower transverse beam (2) Jack with double effect (3.1) Lateral support (3.2) Supports perpendicular to frame plate (4) Load cell (5) Upper transverse beam (6) Support for upper transverse beam (7) Furnace (8) Crossing column (Dotreppe at el., 1996) ................................................................................................................... 74 Figure 2.53. NSC and HSC column after ASTM E119 fire tests (Kodur, 1999) ....... 76 Figure 2.54. Temperature distribution at various depths in NSC and HSC columns (Kodur, 1999)...................................................................................................... 77

xviii Figure 2.55. Axial deformation histories of NSC and HSC columns during fire exposure (Kodur, 1999) ...................................................................................... 78 Figure 2.56. Position of thermo couples and load set-up (di Prisco et al., 2003) ....... 79 Figure 2.57. Comparison between the standard time-temperature curve and the actual heating curves (di Prisco et al, 2003).................................................................. 80 Figure 2.58. Temperature distribution according to depth inside SFRC slab and vertical displacement of SFRC and plain concrete slabs (di Prisco et al, 2003) 81 Figure 2.59. Virgin, hot and residual load-deflection of slabs (di Prisco et al, 2003)82 Figure 2.60. Loading frame and test set-up (Foster et al., 2004)................................ 83 Figure 2.61. Supporting frame and heating elements (Foster et al., 2004)................. 83 Figure 2.62. Locations of loads and displacements gages (Foster et al., 2004).......... 84 Figure 2.63. Temperature profiles (Foster et al., 2004) .............................................. 84 Figure 2.64. Yield line mechanism and crack patterns (Foster et al., 2004) .............. 85 Figure 2.65. Mid-span deflections for the tested slabs (Foster et al., 2004)............... 86 Figure 3.1. Granite sand and gravel ............................................................................ 89 Figure 3.2. Relative humidity of 24 in. and 48 in. cylinder vs. Time .................... 91 Figure 3.3. SHT 75 sensor and EK-H3 data logger .................................................... 92 Figure 3.4. Plastic tube built with the cable and sensor and 24 inch cylindrical specimen embedded a sensor at center ............................................................... 93 Figure 3.5 Equipments and test set up ........................................................................ 93 Figure 3.6. Temp, RH and Strain vs. Time (Aluminum)............................................ 96 Figure 3.7. Temperature and RH inside the chamber vs. Time (No humidity control) ............................................................................................................................. 98

xix Figure 3.8. Temperature, RH and strain history vs. Time (No humidity control) ...... 99 Figure 3.9. RH, temperature, and strain history vs. time (No humidity control)........ 99 Figure 3.10. Part C magnified in Fig. 3.9. ................................................................ 100 Figure 3.11. Dilatation of solid microstructure induced by decrease of capillary tension with temperature increase..................................................................... 103 Figure 3.12. Length change of Portland cement paste specimens at elevated temperature: (a) Philleo (1958); (b) Harada et al. (1972); (c) Cruz and Gillen (1980); Crowley (1956) .................................................................................... 103 Figure 3.13. Linear thermal expansion of various rocks at elevated temperature: (a) sand stone; (b) limestone; (c) granite; (d) anorthosite; (e) basalt; (f) limestone; (g) sandstone; (h) pumice (Soles and Gellers, 1964)............................................. 104 Figure 3.14. Conventional thermal strain by temperature increase from 28 C to 70 C ........................................................................................................................... 105 Figure 3.15. Pure-Hygro Strain (PHS) under constant temperature 77.5 C 0.5 ... 106

Figure 3.16. RH, temperature, and strain history vs. time ........................................ 107 Figure 3.17. Shrinkage strain according to internal RH decrease of concrete under constant temperature 24.5 C............................................................................ 107 Figure 3.18. Temperature and RH inside the chamber vs. Time .............................. 109 Figure 3.19. Temperature, RH and strain history of the specimen vs. Time ............ 109 Figure 3.20. RH, temperature, and strain history vs. time (Dried specimen) ........... 110 Figure 3.21. Pure thermal strain by temperature increase from 28 C to 70 C ....... 110 Figure 3.22. RH (%) and Temp. (Kelvin) versus factor (f) for the range A in Fig. 3.9 ........................................................................................................................... 112

xx Figure 3.23. Equivalent RH (%) at Temp. 77.5C (350.65 Kelvin) versus factor (f) ........................................................................................................................... 112 Figure 3.24. Thermo-hygro coupling strain between 28-70C................................. 113 Figure 4.1. Schematic of temperature scenarios ....................................................... 119 Figure 4.2. Test set-up for heating ............................................................................ 119 Figure 4.3. Temperature histories measured at the inside of the mullite tube .......... 122 Figure 4.4. Specimen geometry, locations of thermocouples and test set-up........... 123 Figure 4.5. Transient temperature at each position................................................... 124 Figure 4.6. T Vs. Surface temperature ................................................................... 125 Figure 4.7. Temperature profiles during heating and cooling .................................. 127 Figure 4.8. Thermal diffusivity................................................................................. 128 Figure 4.9. Procedure of the test sep-up ................................................................... 129 Figure 4.10. Test set-up detail................................................................................... 130 Figure 4.11. Cumulative water permeated versus time for the test series R15_D2.. 131 Figure 4.12. Cumulative water permeated versus time for the test series R2_D4.... 132 Figure 4.13. Linear regression for the test series R15_D2 ....................................... 134 Figure 4.14. Linear regression for the test series R2_D4 ......................................... 134 Figure 4.15. Relative water permeability (test series R15_D2)................................ 136 Figure 4.16. Relative water permeability (test series R2_D4).................................. 137 Figure 4.17. Comparison between R2_D4 and R15_D2 subjected to slow cooling. 138 Figure 4.18. Comparison between R2_D4 and R15_D2 subjected to natural cooling ........................................................................................................................... 138 Figure 4.19. Test set-up of the residual compression test......................................... 139

xxi Figure 4.20. Relative velocity versus maximum temperature exposed .................... 141 Figure 4.21. Relative elastic modulus (from UPV) versus maximum temperature.. 143 Figure 4.22. Stress versus strain (Slow cooling)....................................................... 144 Figure 4.23. Stress versus strain (Natural cooling)................................................... 144 Figure 4.24. Stress versus strain (Water cooling) ..................................................... 145 Figure 4.25. Stress versus strain (Maximum temperature 200C) ............................ 145 Figure 4.26. Stress versus strain (Maximum temperature 400C) ............................ 146 Figure 4.27. Stress versus strain (Maximum temperature 600C) ............................ 146 Figure 4.28. Stress versus strain (Maximum temperature 800C) ............................ 147 Figure 4.29. Relative residual strength vs. maximum temperature .......................... 148 Figure 4.30. Relative initial tangent modulus vs. maximum temperature ................ 150 Figure 4.31. Comparison between E p and Ei (Slow cooling).................................. 151 Figure 4.32. Comparison between E p and Ei (Natural cooling).............................. 151 Figure 4.33. Comparison between E p and Ei (Water cooling) ................................ 152 Figure 4.34. Effect of heating on weight loss (RH 100% and 90%) ........................ 154 Figure 4.35. Specimens exposed to heating rate 2 C/min (Natural cooling)........... 156 Figure 4.36. Specimens exposed to maximum temperature 800 C (Natural cooling) ........................................................................................................................... 156 Figure 4.37. Granite concrete melted in the furnace................................................. 157 Figure 5.1. Multiscale internal structure of concrete ................................................ 161 Figure 5.2. Hydration process of cement paste......................................................... 163 Figure 5.3. Change of phase composition with increasing temperature (w/c=0.67) 176

xxii Figure 5.4. Comparison between current model and test data by Piasta et al. (1984) ........................................................................................................................... 179 Figure 5.5. Comparison between current model and model of Harmathy (1970) .... 179 Figure 5.6. Comparison between current model and test data.................................. 192 Figure 6.1. Simplification to spherical model........................................................... 196 Figure 6.2. Three-phase effective media model........................................................ 197 Figure 6.3. Thermal strain test data of dehydrated substances ................................. 200 Figure 6.4. Thermal strain test data of hydrated substances ..................................... 200 Figure 6.5. Thermal strains of phases in cement paste used in model...................... 204 Figure 6.6. Arrangement of phases in cement paste ................................................. 206 Figure 6.7. Comparison between model and experimental data for cement paste ... 207 Figure 6.8. Arrangement of phases in concrete ........................................................ 208 Figure 6.9. Comparison between model and experimental data (limestone concrete) ........................................................................................................................... 209 Figure 6.10. Comparison between model and experimental data (sandstone concrete) ........................................................................................................................... 210

CHAPTER 1

1.

INTRODUCTION

1.1. Background

Concrete is a low conductor which exhibits high resistance to temperature transients. However, extreme and rapid heating from fire can cause large volume changes due to thermal dilatation, shrinkage due to moisture migration, and eventual spalling due to high thermal stresses and pore pressure build-up. The large volume change produces stresses that result in microcracking and large fractures which may lead to structural failure. The extent of the concrete property variation due to high temperature depends on many internal and external parameters, such as concrete mix design, properties of the constituents, heating rate, cooling rate, maximum exposed temperature, etc. Various studies for concrete exposed to high temperature have been conducted experimentally and analytically on the importance of different material parameters to provide essential information to the concrete industry for improving the fire resistance of concrete. Recently, the research has been concentrated on various coupling effects of the parameters such as thermo-chemo-mechanical coupling and thermo-hygro-mechanical coupling.

2 However, more studies on this important topic are still needed, because the properties of concrete under high temperature have not been fully understood. Even if there are numerous experimental and theoretical results available in the literature, contradictions among observations exist and need to be reconciled.

1.2. Thesis objectives

The first objective of this research is to find the coupling effect of temperature and moisture on deformation of concrete, experimentally. Due to the dependency of the relative humidity on temperature, it is not easy to measure the temperature effect and moisture effect on strain separately over time. Usually, the thermal strain of concrete, which is called Conventional Thermal Strain (CTS) in this study, is obtained by measuring the displacement change without moisture control. In this study, CTS, the strain caused by temperature increase under constant humidity (Pure Thermal Strain: PTS), and the strain caused by moisture change under constant temperature (Pure Hygro Strain: PHS) are measured continually and simultaneously over time. PTS and PHS do not have a coupling effect between temperature and moisture. From the data analysis based on the measured strains, the thermo-hygro coupling effect is obtained. The second objective is to investigate the strength, stiffness, and durability performance of concrete subjected to various heating and cooling scenarios. Extensive experimental studies on this important topic were performed previously. The important experimental parameters included maximum temperature, heating rate,

3 types of aggregates used, various binding materials, and mechanical loads under high temperature conditions. However, the experimental studies have mainly concentrated on the strength of concrete; very little research focused on the reduction of concrete durability caused by fire damage. The unstressed residual compressive strength test and Ultrasonic Pulse Velocity test (UPV) are performed to investigate the strength and stiffness deterioration by each high temperature scenario. The reduction of the durability of concrete is investigated using a water permeability test (WPT). Additionally, weight losses, color changes, and cracks of the specimens are also studied and reported. The third objective is to study mechanical properties of concrete under high temperatures from both mechanical and chemical points of view. The variation of the mechanical properties of concrete results mainly from two mechanisms. One is the variation of material properties of the constituent phases under high temperatures, and the other is the transformation of constituent phases under different temperatures. Despite various experimental studies, prediction models have not yet been developed for the thermal strain and thermal degradation of stiffness of concrete in which phase transformations at the micro-scale level are considered. This study presents multiscale chemo-mechanical models for thermal strain and modulus of elasticity of concrete considering phase transformations due to high temperatures.

1.3. Thesis organization

Chapter 1 covers a brief introduction, thesis objectives, and organization.

4 Chapter 2 presents an extensive literature review on the microstructure variations of concrete subjected to high temperature, the properties of plain concrete tested in high temperature environments, and the behavior of reinforced concrete components such as beams, columns, and slabs under high temperatures. Chapter 3 presents the testing procedure and the test results on Conventional Thermal Strain (CTS), Pure Thermal Strain (PTS), and Pure Hygro Strain (PHS). Also, the thermo-hygro coupling effect from the data analysis based on the measured strains is presented. Chapter 4 presents the testing procedures and results on the strength, stiffness, and durability performance of concrete subjected to various heating and cooling scenarios. Additionally, weight losses, color changes, and cracks of the specimens are also studied and reported. Chapter 5 presents the prediction model for the thermal degradation of stiffness of concrete in which phase transformations at the micro-scale level are considered. Chapter 6 presents the model for the thermal strain of cement paste and concrete considering the phase transformations due to high temperature. Chapter 7 consists of summary and conclusions.

CHAPTER 2

2.

LITERATURE REVIEW

2.1. Microstructure variations of concrete due to high temperature

Material properties (such as strength, stiffness, coefficient of thermal expansion, thermal diffusivity, and moisture diffusivity, etc) of concrete under high temperature are linked with intricate micro-structural variations, such as the transformation of constituent phases. At microstructure level, various experimental studies of cement paste and concrete are summarized in the following sections.

2.1.1. Microstructure of hydrated cement paste

Cement chemists use the following the abbreviations: C = CaO ; S = SiO2 ;

A = Al2O3 ; F = Fe2O3 ; S = SO3 ; H = H 2O . Anhydrous Portland cement is a gray

powder composed of angular particles typically ranging in the size from 1 to 50 m. It is produced by pulverizing the clinker being a heterogeneous mixture of several compounds produced by high-temperature reactions between calcium oxide ( CaO ), silica ( SiO2 ), alumina ( Al2O3 ), and iron oxide ( Fe2O3 ) with a small amount of

6 calcium sulfate (gypsum: CaSO4 .2 H 2O ). The major compounds of Portland cement are C3 S (tricalcium silicate called as alite phase), C2 S (dicalcium silicate called as
belite), C3 A (tricalcium aluminate called as aluminate phase), and C4 AF

(tetricalcium aluminoferrite called as ferrite phase). In ordinary Portland cement, their respective amounts usually range between 45 and 60, 15 and 30, 6 and 12, and 6 and 8 percent. The hydrated cement pastes can be classified as four principal substances described below:
Calcium silicate hydrate (tobermorite gel): The calcium silicate hydrate

phase (C-S-H) makes up 50 percent to 60 percent of the volume of solids in a completely hydrated Portland cement paste. Thus, C-S-H is the most important phase determining the properties of the paste. The term hyphenated of C-S-H indicates that C-S-H is not a well-defined compound. Actually, the ratio of cement to sand (c/s) varies between 1.5 and 2.0, and the structural water content varies even more. The morphology of C-S-H also varies from poorly crystalline fibers to reticular network. In accordance with Diamond (1978), the morphology of C-S-H gels has been divided into 4 types. Type I is fibrous particles that are a few micrometers long in the form of a spine, prism, rod, or rolled sheet. Type II is reticular or honeycomb-shaped structure formed in conjunction with Type I. Type III is the nondescript or flattened particles under 0.1 m. Type IV is compact and dimpled appearance, which is generally formed in a later hydration period (Ramachandran, 1981). Fig. 2.1 shows SEM (Scanning-Electron-Microscopy) image of tobermorite in concrete.

Figure 2.1. SEM image of tobermorite (C-S-H, Type I)

Calcium hydroxide (portlandite): Calcium hydroxide crystals [ Ca (OH ) 2 ]

comprise 20 percent to 25 percent of the volume of solids in the hydrated paste. In contrast to the C-S-H, calcium hydroxide is a compound with a definite stoichiometry. It tends to form large crystals with distinct hexagonal-prism morphology. Later in hydration the morphology usually varies to massive structure that loses its hexagonal form. Fig. 2.2 shows SEM image of portlandite.

Figure 2.2. SEM image of portlandite (CH) crystals

Calcium sulfoaluminates hydrates: Calcium sulfoaluminate hydrates occupy

15 percent to 20 percent of the solids volume in the hydrated paste. Thus, it takes part in only a minor role in the microstructure-property relationships. During the early stages of hydration, the sulfate/alumina ionic ratio of the solution phase is generally the formation of trisulfate hydrate ( C6 AS 3 H 32 ), which also called ettringite. Ettringite forms into long rods or needles with parallel sides that have no branches. In pastes of ordinary portland cement, ettringite eventually transforms to the monosulfate hydrate ( C4 ASH18 ), which forms hexagonal-plate crystals. The presence of the monosulfate hydrate in portland cement concrete makes the vulnerable concrete to sulfate attack. It should be noted that both ettringite and the monosulfate contain small amounts of iron, which can substitute for the aluminum ions in the crystal structure. Fig. 2.3 shows SEM (Scanning-Electron-Microscopy) images of ettringite crystals.

(a) Hexagonal form

(b) Tubular form

Figure 2.3. SEM images of ettringite crystals

9
Unhydrated clinker grains: Depending on the particle size distribution of the

anhydrous cement and the degree of hydration, some unhydrated clinker grains may be found in the microstructure of hydrated cement paste, even long after hydration. As stated earlier, the clinker particles in modern portland cement generally conform to the size range 1 m to 50 m. With the progress of the hydration process, the smaller particles dissolve first and disappear from the system, then the larger particles become smaller. Because of the limited available space between the particles, the hydration products tend to crystallize in close proximity to the hydrating clinker particles, which gives the appearance of a coating formation around them. At later ages, hydration of clinker particles due to the lack of available space results in the formation of a very dense hydration product, the morphology of which may resemble the original clinker particle. Fig. 2.4 shows a model of well-hydrated Portland cement paste.

Figure 2.4. Model of a well-hydrated Portland cement paste

In Fig. 2.4, A represents aggregation of poorly crystalline C-S-H particles which have at least one colloidal dimension (1 m to 100 m). Inter-particle spacing

10 within an aggregation is 0.5 nm to 3.0 nm (average 1.5 nm). H represents hexagonal crystalline products such as CH , C4 ASH18 , C4 AH19 . They form large crystals, typically 1 m wide. C represents capillary cavities or voids which exist when the spaces originally occupied with water do not get completely filed with the hydration products of cement. The size of capillary voids ranges from 10 nm to 1 m, but in well-hydrated pastes with low water ratio to cement, they are less than 100 nm.

2.1.2. Microstructure changes of cement paste due to high temperature

A common link throughout all concrete types is the use of cement paste as a matrix. The response of the cement paste correlated with temperature is essential to understanding characteristics of concrete exposed to high temperature. Piasta et al (1984 b) investigated the microstructure and phase composition of cement pastes from 20-800 C. The reactions of Ca (OH ) 2 , CaCO3 , C-S-H, nonevaporable water and micropores to heat were investigated using thermal analysis (TA), X-ray diffraction analysis, infrared spectroscopy analysis and mercury porosity. Ordinary Portland cement (OPC) specimens (initial w/c=0.4) were cured for 28 days under the curing condition of relative humidity of about 95 percent at temperature 20 2 C . The hardened OPC specimens were dried at temperature 105 5 C and then the specimens were kept for 3 hours in each of the invesgated temperatures, in the range of 200 C to 800 C in intervals of 100 C. Water content in cement paste decreases up to 600 C. Beyond this temperature water content holds fairly consistent with a slight increase. This fact is the same as the results obtained from the

11 measurement for weight loss of the concrete specimens in current research, even though the materials (cement paste in the research of Pista et al. and concrete in current research) and heating conditions are different with those used in the research of Piasta et al (see section 4.7). Water vapor created in the concrete between 100300 C causes high pressure in the paste. This forms most ideal conditions for internal autoclaving because steam is liberated most intensively in this temperature range. In this temperature range, additional hydration of unhydrated cement grains occurs as the result of steam effect under the internal autoclaving. This is manifested by increase of the content of Ca (OH ) 2 that is observable from the TG [Fig. 2.5 (a)] and X-ray method [Fig. 2.5 (b)] and decrease of unhydrated cement grain quantities ( C3 S + -C2 S ) observable from the X-ray method.

(a) Ca (OH ) 2 and CaCO3 (TG)

(b) Ca (OH ) 2 and -C2 S + C3 S (X-ray)

Figure 2.5. Ca (OH ) 2 contents observed from TG and X-ray (Piasta et al, 1984 b)

The difference of TG method (decrease) and X-ray method (increase) for calcium hydroxide between 200 C and 300 C is due to different sensitivities of both

12 methods to cement paste structure. The indications of TG method are not practically dependent on the structure. On the other hand, the X-ray method exhibits higher sensitivity to change in the structure, particularly to the degree of its crystallization. Keeping this in mind, up to 300 C progressing recrystallization of amorphous

Ca (OH ) 2 by the internal autoclaving conditions occurs in the hardened cement paste.
The results observed from TG method also show that the carbonization kinetics of Ca (OH ) 2 in the temperature range of 200-500 C increases as indicated by a decrease in Ca (OH ) 2 in favor of CaCO3 . Ca (OH ) 2 , which has not undergone carbonization, is decomposed with the emission of calcium oxide between 450 C and 550 C. In the temperature range of 500-800 C, the increase in the Ca (OH ) 2 content is a result of humidity effect from air on free lime CaO being formed in this temperature range. The content of CaCO3 decreases gradually as a result of its dissociation at temperature above 500 C. In opinion of Tsivilis et al (1998), CaCO3 begin to decompose in the temperature range of 600-800 C (decaonation of calcite) and thus increases the content of calcium oxide ( CaO ).

Figure 2.6. SEM images of calcite crystals

13
2.1.3. Porosity and pore size distribution according to temperature

Table 2.1 shows the results of porosity tests between 20-800 C. Total porosity at room temperature is the lowest.

Table. 2.1. Porosity and Pore Size Distribution (Piasta et al, 1984 b)

At 200 C porosity is not significantly affected in either total porosity or pore size distribution. Between 300-500 C, total porosity along with simultaneous slight changes in mercury porosity increases significantly due to an increase in the percentage of pores greater than 7500 nm in diameter. In mercury porosity, there is a decrease in the 5-10 nm pores with 25-75 nm pores increasing. The formation of microcracks would cause the increase in pores with diameters greater than 500 nm in the temperature range of 300-500 C. At 600 C an almost doubling of the total porosity occurs. Also at this temperature capillary pores, 250-500 nm diameters, are significantly increased. This rapid increase in porosity is caused by two chemical

14 processes, liberation of water from the decomposition of

Ca (OH ) 2

[ Ca (OH ) 2 CaO + H 2O ] and the liberation of CO2 resulting from CaCO3 dissociation [ CaCO3 CaO + CO2 ]. 700 C shows the highest total porosity. The percentage of pores in the 25-75 nm diameter range decrease with capillary pores in the range of 250-1000 nm increasing. Total porosity decreases at 800 C. 5-10 nm pores are significantly reduced with 500-1000 nm pores increasing significantly. The beginning of cement paste sintering may cause this. Table 2.2 shows a summary of temperature regimes of chemical reactions of hydrate phase of hardened cement paste according to Schneider and Herbst (1989).

Table 2.2. Temperature regimes of chemical reactions of hydrate phase of hardened cement paste according to Schneider and Herbst (1989)

15 But, the temperature range of decomposition of CaCO3 (calcite) is different from the results of Lach (1970) and Tsivilis et al (1998). In their results, the decomposition of calcite was occurred in temperature range of 600-800 C. The result is very similar to that of Piasta et al (1984b). In results of Piasta et al, the decomposition of calcite was started above 500 C.

2.1.4. Heat deformation of cement paste due to high temperature

One of the destructive causes to concrete failure is the heat deformation in the micro and macro components. Due to differential thermal expansions of phases in hardened cement paste, micro-cracking occurs on the phase boundaries. This microcracking leads to significant destruction of the concrete structure. Piasta (1984a) conducted experiments under the purpose of defining the character and course of heat deformation of phases present in hardened cement paste, and to determine initial temperatures, in temperature range of 20-800 C. The thermal analysis was performed with differential thermal analysis (DTA), thermal dilatometry (TD), and differential thermal dilatometric analysis (DTD). The TD tracked heat deformations while the DTD tracked the velocity of changes and the DTA observes the kind of chemical processes occurring. Portland cement as well as C3S , -C2 S ,

C3 A , and C4 AF which are the mineralogical composition of Portland cement were

studied in both the unhydrated and hydrated states. The mineralogical composition of Portland cement used in the experiments was the followings: C3S -63.2 %, -C2 S 15.4 %, C3 A -9.9 %, and C4 AF -8.1 %. The contents of free calcium oxide ( CaO ),

16 magnesium oxide ( MgO ), alkali ( Na2O + K 2O ), and gypsum were 1.1, 0.8, 0.5, and 3.4 % respectively. A heating rate10 C/min was used.

Figure 2.7. Thermograms of unhydrated clinker minerals and Portland cement: 1) C3S , 2) -C2 S , 3) C3 A , 4) C4 AF , 5) Portland cement

In the temperature range of 20-900 C, the test results from DTD (heat deformation intensity) and DTA (endothermal phenomenon) showed no chemical processes of C3S , C3 A , and C4 AF , whereas -C2 S showed a weak endothermal phenomenon and a strong heat deformation intensity between 630-700 C due to a polymorphous transformation of -C2 S into ' -C2 S (see Fig. 2.7). Thermal

17 expansions of the clinker minerals were tracked with temperature. All minerals showed a linear expansion except for -C2 S , which shows a nonlinear segment between 600-700 C. This expansion is also connected to the polymorphous transformation. Thermal expansion of C3 A was the lowest followed by C3S and C4 AF . Portland cement also showed a linear thermal expansion that was higher than all but -C2 S .

Figure 2.8. Thermograms of hydrated materials: 1) C3S H , 2) -C2 S H , 3) C3 AH , 4) C4 AFH , 5) Portland cement paste, 6) Ca (OH ) 2 , 7) Ettringite

18

Hydrated clinker materials and cement paste has performed differently from that of unhydrated materials. From TD analysis, the hydrated materials between 20200 C showed a low volume expansion similar to that of the unhydrated materials (see Fig. 2.8). Beyond 200 C they start to contract at varying intensities. -C2 S H showed the least contraction followed by C4 AFH , C3S H and C3 AH . Shrinkage continues up to the temperature of dehydration decay for hydrates. After dehydration, hydrates expand with rising temperature. On the other hand, Ca (OH ) 2 had the greatest expansion that lasted from 200 C to 450 C. At 450 C decomposition begins and Ca (OH ) 2 begins to shrink. Ettringite had the shortest and least thermal expansion up to 50 C. However, Ettringite had the greatest shrinkage of all materials tested. All materials showed non-linear heat deformations with a correlation in the heat deformation intensity (DTD) and the endothermal phenomena (DTA) concluding the influence of chemical processes on the heat deformations. The common trend of all materials is the initial expansion up until the onset of dehydration. The dehydration causes shrinkage of the hydrated materials until full dehydration is reached. After dehydration, hydrates expand again up to 800 C. 150 C is the point at which shrinkage of the hydrated materials takes a significant role in the thermal deformations. At 550-600 C the hydrated materials are fully dehydrated. After that, the materials are expanded with temperature increase. The diverse nature of concrete microstructure leads to thermal deformations in both the hydrated and unhydrated components. As mentioned in the heat deformations of the unhydrated state, the unhydrated materials were subjected to thermal expansion over

19 all temperature range. Expansion in the unhydrated components with a corresponding shrinkage in the hydrated components leads to stress concentrations along the boundaries. The stress concentrations are linked with microcracks. In SEM observation of Piasta (1984a), micro-cracks first appeared at 300 C around Ca (OH ) 2 . The cracks are linked with the different heat deformations between Ca (OH ) 2 and other materials. A great number of micro-cracks were observed to between large unhydrated particles and the cement paste at 400 C. The cracks are linked with the stress concentrations. At 500 C the interior of unhydrated clinker was completely destroyed. The rehydration of free lime ( CaO ), which has 44 % volume expansion, is correlated to severe cracking due to rapid cooling. Table 2.3 is a summary for heat deformation of cement paste.

Figure 2.9. 1) Ca (OH ) 2 , 2) cement gel [300 C exposure (1000), Piasta (1984a)]

20

Figure 2.10. Destroyed grains of unhydrated clinker as a result of different thermal deformations [500 C exposure (3000), Piasta (1984a)]

21

Table 2.3. A summary for heat deformation of cement paste Temperature 20-900 C 630-700 C Heat deformation & chemical reaction (Unhydrated substances): C3S , -C2 S , C3 A , C4 AF , and Portland cement C3S , C3 A , C4 AF : No chemical reaction Polymorphous transformation of -C2 S into ' -C2 S Thermal expansion (No contraction): -C 2 S > Potland cement > C4 AF > C3 S > C3 A
* Portland cement, C4 AF , C3S , C3 A : Linear thermal expansion * -C2 S : Nonlinear thermal expansion in temperature range 600- 700 C due to a polymorphous transformation of ' -C2 S Temperature 20-900 C 20-200 C 20-600 C Heat deformation & chemical reaction (Hydrated substances): C3S H , -C2 S H , C3 AH , C4 AFH , Portland cement paste, Ca (OH ) 2 , and Ettringite All materials: chemical reaction and nonlinear heat deformation Low thermal expansion of all hydrated materials 1. Shrinkage of the materials due to dehydration of hydrates (except for Ca (OH ) 2 ) 2. Ca (OH ) 2 : Greatest thermal expansion up to 450 C Occurrence of initial micro-cracks due to different heat deformation between Ca (OH ) 2 (expansion) and other materials (shrinkage) Occurrence of a great number of micro-cracks between large unhydrated particles and the cement paste * Due to stress concentrations along the boundaries by different heat deformations between the unhydrated components (expansion) and hydrated components (shrinkage) Interior of unhydrated clinker is completely destroyed The hydrated materials are fully dehydrated. After full dehydration of hydrates, they expand again up to 800 C Rehydration of free lime ( CaO ): CaO + H 2O ( From air or water ) Ca(OH ) 2 44 % volume expansion Correlation to severe cracking

20-900 C

300 C

400 C

500 C 550-600 C 600-800 C Cooling

22
2.1.5. SEM (Scanning Electron Microscopy) observation of concrete

In a study conducted by Lin et al. (1996), the microstructure of concrete exposed to elevated temperatures in both actual fire and laboratory conditions were evaluated with the use of SEM and stereo microscopy. Thermal stresses, due to thermal gradients, decomposition of calcium hydroxide, calcinations of the limestone aggregates and phase transformation of quartz aggregate, cause spalling and cracking during heating and disintegration during cooling. When temperatures rise about 100 C evaporable moisture reduces cohesive forces between C-S-H layers and gel surface. In samples exposed to temperatures below 200 C no noticeable cracks were observed. Exposure to 400 C causes the collapse of crystals. Buildup of internal pressure is caused by vaporization of free water, dehydration of calcium hydroxide, which occurs exclusively between 440-580 C, above 350 C, and partial volatilization of C-S-H gels above 500 C. At 573 C the crystal structure of quartz in siliceous aggregate transforms from low temperature crystal to high temperature

crystal.

Figure 2.11. Morphologies of hydrates after 400 C exposure (Lin et al, 1996)

23

Fig. 2.11 shows the morphology of hydrates after an exposure to 400 C. Above 350 C calcium hydroxides dissociates into lime ( CaO ) and water. This leads to damage as free lime expands during cooling. Dehydroxylation of calcium hydroxide occurs exclusively between 440 C and 580 C (above 350 C).

Figure 2.12. Morphologies of hydrates after 500 C exposure (Lin et al, 1996)

. Figure 2.13. Crakes and honey combs after 900 C exposure (Lin et al, 1996)

24 Fig. 2.12 shows the hydrates after 500 C exposure. Calcium hydroxide appears layered plates with ettringite and C-S-H from about 400 C. Fig. 2.13 also shows the layered plates. Significant shrinkages, cracks and honeycombs were observed in concrete samples exposured to 900 C. Fig. 2.14 shows the morphology of a sample after 900 C exposure. After the sample was heated to 900 C, the sample under water-curing formed a different microstructure as compared with compared with that under air-curing. This is due to the presence of water. With the presence of water, CH is formed as a result of rehydration of calcium oxide. The C-S-H forms honeycomb and filigree networks with small pore spaces.

Figure 2.14. Sample under water cooling after 900 C exposure (Lin et al, 1996)

Wang et al. (2005) used SEM to examine the cracking of high performance concrete (HPC) exposed to high temperatures under axial compressive loading of about 200 N. The compressive loading mainly restricted the loose deformation of the sample at high temperature. The study observed the initiation and propagation of cracks and micro-cracks in HPC. Temperatures were increased from 25 C to 100,

25 200, 300, 400, 500 C. The interfacial transition zone (ITZ) has higher porosity than the surrounding matrix. Micro-cracking is affected not only by the applied stress but also by the exposure temperature. Cracks formed around the aggregate become significant as temperature increases. This is most likely due to the changed properties along the ITZ and dilation differentials. Micro-cracks formed in the direction of the applied load. Between 100 C and 200 C micro-cracking was initiated in loading direction although this did not lead to the cleavage failure of aggregate particles. However micro-cracking did not continue to propagate in this direction from 300500 C. In this temperature range the combination of C-S-H becoming soft and increased adhesion between intermixed and hydrated phases begins to close cracks. Therefore 200 C is a critical temperature for the formation of micro-cracks of HPC under service conditions.

2.1.6. SEM observation of concrete used in current research

SEM observation was conducted from the concrete sample used in the current research. The average size of samples used was 15-20 mm in diameter. The samples were heated with rate of 2 C/min. After heating up to the desired temperature, the specimens were cooled naturally inside the closed furnace. The exposed maximum temperatures were 24, 500, 700, and 900 C. Fig. 2.15 (a) shows hydration products of calcium silicate gel (C-H-S), CH, ettringite in unheated sample. The ettringite are in the form of rods interspersed between hydrates. Fig. 2.15 (b) shows well-developed ettringite and micro-cracks developing between calcium hydroxide in the cement

26 paste. Fig. 2.15 (c) shows the morphologies of C-S-H and CH, and ettringite after exposure 700 C. Fig. 2.15 (d) shows severe shrinkage cracks and honeycombs in cement paste after exposure at 900 C. The results of the SEM are relatively similar to those from literature review.

Ett CH

CH CH CSH

CH

(a) 24 C exposure (3000)

(b) 500 C exposure (2500)

Shrinkage Cracks

CSH

Honey combs (CSH)

Ett

CH

(c) 700 C exposure (5000)

(d) 900 C exposure (2400)

Figure 2.15. SEM images of the concrete used in current research

27
2.2. Pure concrete exposed to high temperature

2.2.1. Two major problems when concrete is exposed to fire

Fire incurs two major problems in concrete. One is the deterioration in mechanical properties of concrete, such as physicochemical changes of the cement paste and aggregate, thermal incompatibility between the aggregate and the cement paste according to temperature level, heating rate, applied load, and moisture loss. The other problem is spalling of concrete. Spalling can lead to severe reduction of the cross sectional area, which leads to the exposure of the reinforcing steel to excessive temperatures. Fig. 2.16 illustrates the physicochemical processes taking place in concrete at various temperature ranges. A detailed review of concrete spalling and the physical mechanisms behind this phenomenon is presented in Sec. 2.2.2.

Figure 2.16. Physicochemical process in Portland concrete during (Khoury, 2000)

28
2.2.2. Selection of concrete mix

Selection of the concrete mix design is very important in the fire design for reinforced concrete structures. The individual material constituents should be selected considering spalling and strength loss. Fig. 2.17 presents a schematic for the two arguments of appropriate material choice.

Figure 2.17. Material choices for concrete under high temperature (Khoury, 2000)

a. Mix designs considering strength loss

The mechanical properties of concrete under high temperature can be improved significantly by intelligent mix design. There are three items to be considered: aggregate, cement paste and the interaction between them. The choice of aggregate is very important in concrete mix design since its thermal stability depends

29 strongly on the type of aggregate being used. In general, flint gravel breaks up at relatively low temperatures (below 350 C), while granite exhibits thermal stability up to 600 C. Other important features of the aggregate are low thermal expansion, rough angular surfaces and the presence of reactive silica. Under high temperature, aggregates expand and cement paste shrinks due to rapid moisture loss which cause crack on the interface between aggregate and paste. Therefore, aggregates with low coefficient of thermal expansion can reduce the thermal incompatibility between cement paste and aggregate, hence use of low coefficient of thermal expansion reduces the thermal damage. Also, rough and angular surface of the aggregate improves the physical bond with the cement paste. An important fire design parameter is the cement/sand (c/s) ratio. A low c/s ratio results in a low calcium hydroxide (Ca(OH)2) content in the original mix and ensures beneficial hydrothermal reaction. Calcium hydroxide is not desirable because it dissociates at about 400 C into CaO and CO2 . CaO rehydrates expansively and detrimentally upon cooling and exposure to moisture. So, the reduction of C/S ratio using blast furnace slag, fly ash, and silica fume can lead to an improvement of concrete strength upon fire. Tests by Khoury at el., (1995 (a)) show that the use of slag produces the best results at high temperature, followed by fly ash and then the silica fume. Particularly, the relatively poor performance of the silica fume cement paste (contrary to its high durability at room temperature) may be attributed to the dense, low permeability structure of the paste which does not readily allow moisture to escape from heated concrete, which results in high pore pressures and occurrence of micro-cracks.

30
b. Mix designs considering spalling

Spalling is the break off of layers or pieces of concrete from the surface of a structural element. The spalling of normal strength concrete (NSC) occurs due to rapid temperature change - typically 20 C/min (Khoury, 2000). High strength concrete (HSC) has a significantly higher potential for explosive spalling than normal strength concrete (NSC) due to its low permeability. Explosive spalling of HSC may occur even at relatively low heating rate - less than 5 C/min (Phan, 2002). However, spalling occurs only in narrow regions of the concrete specimen, which has been observed by many researchers. There has been no explanation as regards to why spalling does not occur in all specimens. Many researchers have been arguing about what is the main cause of explosive spalling.

Table 2.4. Factors influencing spalling of concrete (Khoury, 2000)

Spalling of concrete can be classified into four categories. They are aggregate spalling, surface spalling, explosive spalling as violent breaking and corner spalling as non-violent breaking. The main factor responsible for the first three types is the

31 heating rate, while the fourth type is influenced more by the maximum temperature. The following table summarizes the natures and the main influential factors for concrete spalling. The main factors are heating rate, permeability of concrete, moisture content, presence of reinforcement and level of externally applied load. In order to prevent the occurrence of concrete spalling, it is very important to understand what happens in concrete that causes spalling, that is, to understand the fundamental mechanisms that cause concrete spalling. There are several theories explaining the spalling mechanisms, which may be classified in three categories: (a) pore pressure spalling, (b) thermal stress spalling, and (c) combined pore pressure and thermal stress spalling.
- Pore pressure spalling: Fig. 2.18 shows the mechanism of pore pressure spalling.

High temperature causes the evaporation of free water near the concrete surface. The high vapor pressure in the surface layer drives the water vapor to diffuse in two opposite directions: to the surface and into the deeper part of the concrete specimen. With a sharp temperature increase at the concrete surface (under rapid heating), the interior temperature of concrete remains low. When the free water evaporates the vapor diffuses into the interior part of the concrete (cooler part), where it condensates. The condensation of vapor increases the moisture content of the concrete in that layer and thus reduces the permeability of the concrete, which results in the formation of a barrier in the interior, the so-called moisture clog (see Figs. 2.18 (b) and 2.18 (c)). The interior water vapor is blocked by the clog, and the vapor pressure starts to build up rapidly. As soon as the pressure exceeds the tensile strength, then spalling takes

32 place. Fig. 2.19 shows the conceptual distribution of temperature, pore pressure and moisture in a massive concrete section heated at the unsealed surface on the right side.

Figure 2.18. Mechanism of pore pressure spalling (Schneider and Horbst, 2002)

Figure 2.19. Gradients of temperature, pore pressure and moisture in a massive concrete section heated at the unsealed left-hand surface (Khoury, 2000)

33 In this process, the sharp temperature reduction (high temperature gradient) from the surface plays an important role, and the high gradient occurs only under very rapid heating of massive concrete components. This is why fast heating is a necessary condition for the spalling. Other necessary conditions are low permeability of concrete and the size of the concrete structure, otherwise the vapor would readily escape to the surface and there would be no pressure build-up. On the other hand, high temperature causes the dehydration of chemically bonded water in the cement paste, which contributes to the high pore pressure and spalling. This is why spalling takes place in the HPC concretes with low moisture content. This also explains that high initial water content in concrete is not a necessary condition for spalling.
- Thermal stress spalling: This mechanism is the result of thermo-mechanical

coupling, in which the temperature gradient upon rapid heating causes severe thermal stress gradients in the concrete component. In the high temperature zone (on the surface) concrete expands more than in the low temperature zone (the interior part). As a self-equilibrating thermal stress state develops, a thin layer near the surface is in compression while the interior part is in tension. Because of the high temperature gradient, the compressive stress in the thin surface layer can be very high, which causes buckling and delamination of the outer layer, observed in the form of spalling. Therefore, the main factor of thermal stress spalling is the excessive thermal stress generated by rapid heating of massive concrete components and structures.
- Combined pore pressure and thermal stress spalling: In most cases, a

combination of the two mechanisms takes place. Fig 2.20 shows forces acting in heated concrete.

34

Figure 2.20. Forces acting in heated concrete (Zhukov, 1975)

Explosive spalling generally occurs under the combined effect of pore pressure, and compression in the exposed surface region induced by thermal stress and external loading and internal cracking. Consequently the pore pressure needs to be considered together with both the thermal and the load-induced stresses before the occurrence of explosive spalling.

Figure 2.21. Mechanical property test setup (Phan, 2002)

35

Figure 2.22. Test set up, Pore pressure specimen, and Specimen failure (Phan, 2002)

Fig. 2.21 and Fig. 2.22 show the experimental setup for testing concrete spalling and the fractured concrete specimens at NIST. The methods of improving concrete mix design to prevent spalling will be introduced in later sections.

2.2.3. Testing methods for mechanical properties of concrete

Idealized testing methods: There are six idealized test methods for testing

concrete at high temperature. Four of which belong to the category of steady state,

36 isothermal temperature experiments, and two belong to the category of the transient, isotonic temperature experiments. Steady state tests include stress rate or strain rate controlled tests, steady state creep and relaxation tests. Transient tests include transient creep and relaxation experiments (Phan, 1996).
- Steady state tests: The difference between strain and stress controlled tests depends

on the load control method used in the experiment. The test specimen is heated up to the target temperature with a constant heating rate. After the concrete specimen has reached a uniform temperature distribution, the specimen is subjected to a constant rate of stress or strain until the ultimate stress level or strain level is reached. This test method is also called unstressed test, because the specimen is stress free (or strain free) during the heating period. Data from this type of test can be used to determine the compressive strength, the modulus of elasticity, the strain at ultimate strength, and the dissipated mechanical energy as function of temperatures. The steady creep test is designed to measure creep deformations at different target temperatures (isothermal creep test). The test specimen is slowly heated to the target temperature until a uniform temperature distribution is reached in the concrete specimen. After that the mechanical load is applied. The loading period is typically much longer than the loading period of stress rate controlled or the strain rate controlled experiments. Once the load level is reached, both the target temperature and the mechanical load are kept constant during the test period. The measured results are creep deformations due to sustained constant load at different temperatures. The elastic deformation which occurs instantaneously upon loading is separated from the creep deformations which result from long term, sustained loading. This type of test is not applicable to a

37 concrete structure under fire since the testing duration is normally far longer than the duration of building fires. The steady relaxation test is similar to the steady state creep test. Under each steady temperature condition, an initial (elastic) strain resulting instantaneously from the applied deformation is recorded. Thereafter the initial deformation and the target temperature are kept constant during the testing period and measurements of stress as a function of time are recorded. The steady relaxation test also has little relevance to the performance of concrete under fire situation because its duration exceeds the duration of building fires in real life (Phan, 1996).
- Transient tests: In the case of a transient creep test, the specimen is subjected to a

constant applied load, usually a percentage of the specimens ultimate strength measured at room temperature before heating, and then the specimen is heated with a constant heating rate until failure occurs. In case of a transient relaxation test, the specimen is subjected to a constant applied strain measured at room temperature. The strain is maintained for the duration of the test by adjusting the applied load (usually by reducing the applied load in order to keep the constant strain), while the specimen is heated at a specified rate. The test is terminated when the applied stress level diminishes to zero. Unlike the steady state creep and relaxation tests, where the test durations far exceed the practical duration of building fires, the transient creep and the transient relaxation tests simulate the transient conditions which concrete members might experience in real structures. Thus data obtained from the transient tests have relevance to the performance of concrete structures during fires (Phan, 1996), see also the workshop papers (Willam et al., 2003; Willam et al., 2004).

38
Commonly used test methods: Three testing methods are commonly referred to

as stressed tests, unstressed tests, and unstressed residual strength tests. The schematic of the three test methods is shown in Fig. 2.23.

Figure 2.23. Schematic of temperature and loading histories for the three test methods

Stressed tests are a modified version of stress or strain controlled experiments performed under isothermal temperature conditions. A preload, generally in the range of 20 percent to 40 percent of the ultimate compression strength at room temperature, is applied to the concrete specimen prior to heating, and the load is sustained during the heating period. After the specimen reaches a steady state temperature condition, the stress or the strain is increased with a prescribed loading rate until the specimen fails. The test results of this test are most suitable for representing fire performance of concrete in a column or in the compression zone of beams and slabs. Unstressed tests are carried out identically to the stress or strain controlled experiments of the steady state type. The test results are most suitable for representing

39 the performance of concrete in the tension zone of beam and slabs or concrete elements with low stress levels under service conditions. Unstressed residual strength tests are experiments where the specimen is first cooled to room temperature after one or several cycles of heating without preloading. The load is then applied at room temperature under stress or strain control until the specimen fails. The results of this test are most suitable for assessing the post fire (or residual) properties of concrete (Phan, 1996).

2.2.4. Concrete properties under high temperatures

The compressive strength of concrete at high temperature is largely affected by the following factors: 1) Individual constituent of concrete, 2) Sealing/moisture condition, 3) Loading level during heating period, 4) Testing under hot or cold residual conditions, 5) Rate of heating or cooling, 6) Duration at constant temperature, 7) Time maintained in moist conditions after cooling before the test is carried out, and 8) Number of thermal cycles (Khoury, 2002). In this section, we will focus on the general trends of the test data without giving a detailed description of concrete mix design, curing time and experimental conditions. The effect of elevated temperatures on concrete strength and load-deformation behavior of HSC and NSC were investigated by Castillo and Durani (1990). Type Portland cement with natural river sand and crushed limestone were used for preparing the concrete specimens in the form of 51 mm102 mm cylinders. Fig. 2.24 shows the test results. In the case of the stressed experiment, 40 percent of the

40 ultimate compressive strength at room temperature was applied to the specimens and sustained during the heating period. In the unstressed experiment, when exposed to temperatures in the range of 100 C to 300 C, HSC showed a 15 percent to 20 percent loss of compressive strength, whereas the NSC showed no strength loss (Figure 2.24 (a)).

(a) Compressive strength vs. Temperature (b) Modulus of Elasticity vs. Temperature

(c) Load-deformation of HST

(d) Load-deformation of NSC

Figure 2.24. Hot isothermal test results obtained by Castillo and Durani (1990)

HSC recovered its strength between 300 C and 400 C, reaching a maximum value of 8 percent to 13 percent above room temperature. At temperature above

41 400 C, HSC progressively lost its compressive strength which dropped to about 30 percent of the room temperature strength at 800 C. The trend of NSC was also similar to that of HSC. The elastic modulus in the range of 100 C to 300 C was decreased by 5 percent to 10 percent for both HSC and NSC (Fig. 2.24 (b)). At 800 C, the elastic modulus was only 20 percent to 25 percent of the value at room temperature. Beyond 300 C, the elastic modulus decreased at a faster rate with increase in temperature. The load-deformation plots for HSC and NSC are shown at Fig. 2.24 (c) and (d). NSC specimens did exhibit ductile failure except for 200 C. Between 300 C and 800 C, the NSC specimens were able to undergo large postpeak strains while the decrease in strength was more gradual. HSC showed brittle failure up to 300 C, and with further increasing temperature, the HSC specimens began to exhibit a more gradual failure. Abrams (1971) conducted a study of four variables including aggregate types (carbonate dolomite sand and gravel, siliceous, and expanded shale lightweight aggregates), testing methods (unstressed, stressed and unstressed residual experiments), concrete strengths (ranging from 22.8 MPa to 44.8 MPa), and temperatures (from 93 C to 871 C). The specimens were 75 mm150 mm cylinders. Fig. 2.25 summarizes the test results. The following conclusions were reported: Up to about 480 C, all three concretes exhibited similar strength loss characteristics under each test condition (stressed, unstressed, and unstressed residual). Above 480 C, the siliceous aggregate concrete had greater strength loss and retained less strength for all three test conditions. Specimens made of carbonate aggregates and lightweight aggregates behaved about the same over the entire temperature range and retained

42 more than 75 percent of their original strength at temperatures up to 649 C in unstressed tests.

(a) Carbonate aggregate concrete

(b) Siliceous aggregate concrete

(c) Lightweight aggregate concrete

(d) Stressed tests

(e) Unstressed tests

(f) Unstressed residual tests

Figure 2.25. Hot isothermal test results obtained by Abrams (1971)

43 For the siliceous aggregate, the strength was 75 percent of the original strength at 430 C. Compressive strengths of specimens with preload (stressed tests) were generally 5 percent to 25 percent higher than those without preload (unstressed tests). Also the preloads of 25, 40, and 55 percent of the room temperature compressive strength, had insignificant effect on compressive strength of the stressed specimens. The unstressed residual specimens had the lowest strength compared with the stressed and unstressed specimens tested at high temperatures. The test results of Abrams (1971) indicate that strength recovery took place only in a limited temperature range in the case of the stressed and unstressed experiments.

(a) Residual compressive strength

(b) Residual normalized strength

(c) Residual elastic modulus

(d) Residual normalized elastic modulus

Figure 2.26. Hot isothermal test results obtained by Morita et al. (1992)

44

Morita et al. (1992) conducted unstressed residual strength tests. The specimens were 100 mm200 mm cylinders. The heating and cooling rate were 1 C/min and target temperatures were 200, 350, and 500 C. The keeping time at target temperatures to allow a steady state was 60 min. Fig. 2.26 shows the test results. High strength concrete has a higher rate of reduction in residual compressive strength and modulus of elasticity than normal strength concrete after being exposed to temperatures up to 500 C. The study did not report any spalling problems during heating.

(a) Compressive strength

(b) Elastic modulus

(c) Stress-strain curves (FR-42)

(d) Stress-strain curves at 300C

Figure 2.27. Hot isothermal test results obtained by Furumura et al. (1995)

45

Furumura et al. (1995) performed unstressed tests and unstressed residual tests on 50 mm100 mm concrete cylinders using three compressive strength levels: 21 MPa (normal strength concrete FR-21), 42 MPa (intermediate strength concrete FR-42), and 60 MPa (high strength concrete FR-60). The heating rate was 1 C/min and target temperatures were from 100 C to 700 C with an increment of 100 C. The time at target temperatures to allow a steady state was two hours. The concrete was made from ordinary Portland cement. They observed that, for the unstressed tests, the compressive strength decreased at 100 C, recovered to room temperature strength at 200 C and then decreased monotonically with increasing temperature beyond 200 C. For the unstressed residual tests, the compressive strength decreased gradually with increasing temperature for the entire temperature range without any recovery. The modulus of elasticity, in general, decreased gradually with increase of temperature. Fig. 2.27 shows the test results. As expected, the stress-strain curves for HSC are different from those of normal strength concrete (Fig. 2.27 (d)). High strength concretes exhibited steeper slopes than the NSC at temperature up to 300 C to 400 C in the unstressed test. Noumowe et al (1996) conducted unstressed residual strength tests to compare the performance of HSC exposed to high temperatures with NSC. The specimens were 160 mm320 mm cylinders and 100 mm100 mm400 mm prisms. A normal strength (38.1 MPa) and high strength (61.1 MPa) were used. The prismatic specimens had enlarged ends and were used to measure tensile strength. Calcareous aggregates were used for both concretes. The specimens were heated at a rate of

46 1 C/min to target temperatures of 150, 300, 450, 500, and 600 C, which was maintained for 1 hour, and then allowed to cool at 1 C/min to room temperature.

(a) Residual tensile strengths

(b) Elastic Modulus of HSC and NSC

(c) Porosity vs. temperature

(d) Mass loss vs. temperature

Figure 2.28. Hot isothermal test results obtained by Noumowe et al. (1996)

Uniaxial compressive, splitting tensile, and direct tensile tests were performed to obtain residual compressive strength, modulus of elasticity, and residual tensile strength versus temperature relationships. Figs. 2.28 (a) and (b) show the latter two relationships. Residual tensile strengths for NSC and HSC decreased similarly and almost linearly with increase of temperature. Tensile strengths of HSC at all

47 temperatures were 15 percent higher than those of NSC. Also, the tensile strengths measured by splitting tension experiments were higher than those obtained in direct tension. The residual modulus of elasticity remained approximately 10 percent to 25 percent higher than those of NSC for the entire temperature range. Measurements of porosity after exposure to different temperatures were performed for both concretes using a mercury porosimeter. Fig. 2.28 (c) shows the results of porosity measurements. Fig. 2.28 (d) illustrates the percentage loss in mass for different temperature. Porosity measurements indicated that between 25 C and 120 C, the porosity of both concretes was not altered. As the temperatures increased, NSC became increasingly more porous compared with HSC. Mass losses in both concretes were also similar up to 110 C. The highest rate of mass loss occurred in the temperature range of 110 C to 350 C. The rate of weight loss stabilized at temperatures above 350 C. At any temperature, mass loss in NSC was higher than that in HSC.

2.2.5. The effects of sustained loading during heating

Preloading during heating has positive effects on both the compressive strength and the elastic modulus. Fig. 2.29 demonstrates the positive effects (measured in the hot state for unsealed CRT HITECO ultra-high performance concrete). Comparison of Figs. 2.29 (a) and (b) shows that the compressive strength and elastic modulus of the specimen under sustained loading is larger than those of specimens without the sustained loading. This aspect can be explained from the fact

48 that compressive preloading inhibits crack development - although this explanation has not been fully validated. Mechanical properties such as strength and stiffness generally decrease with increasing temperature. Each of Figs. 2.29 (a) and (b) shows that the compressive strength as well as the elastic modulus decrease as the temperature increases. In other words, the elevated temperature has a very significant effect on the degradation of the mechanical concrete properties.

(a) For 0 % load

(b) For 20 % load

Figure 2.29. The effects of loading and temperature during heating in uniaxial compression of unsealed concrete specimens (Khoury, 2002)

Fig. 2.30 (Khoury, 2002) shows the positive effect of temperature upon the residual (after cooling) compressive strength and the elastic modulus of unsealed C70 HITECO concrete containing thermally stable Gabbro Finnish aggregate. The specimens are heated with heating rate 2 C/min under 0 percent and 20 percent

49 preloaded in compression. The results are shown in terms of percentage of strength and elastic modulus prior to heating.

Figure 2.30. Effect of temperature upon the residual (after cooling) compressive strength and elastic modulus of unsealed C70 HITECO Concrete -20 percent load: expressed as a percentage of strength prior to heating (Khoury, 2002)

2.2.6. Combined thermal and mechanical loading histories

Anderberg and Thelandersson (1976) conducted experiments to determine thermal and mechanical interaction effects. Fig. 2.31 shows the heating scenario and the results of the experiment. The uniaxial strains were measured in two unsealed concrete specimens. In the first case (curve 1 in Fig. 2.31), a uniaxial compressive stress was applied after heating was completed, while in the second case (curve 2 in

50 Fig. 2.31), the same stress was applied from beginning. Fig. 2.31 shows that the total strain (the final part of curve 1) can not be regarded as simple summation of the curves from the stress test and temperature test. It could be argued that the difference between point A and B is caused by creep in the loaded specimen.

Figure 2.31. Measured total strain in concrete specimens heated to 400 C. Applied stress is 45 % of compressive strength at 20 C (Anderberg and Thelandersson, 1976)

But tests on the same concrete under isothermal conditions show very clearly that creep strains in the temperature range 0 - 400 C are much smaller than the observed difference in Fig. 2.31. Therefore, if the observed path dependence is to be interpreted as creep, one must postulate that the creep strain depends on the rate of temperature change. The phenomenon is called transitional thermal creep, a term which was originally introduced by Illston and Sanders (1973). The notation of creep

51 in this context is confusing, since transitional thermal creep is not a long-term effect. In relation to the time scale of isothermal creep it can be regarded as quasiinstantaneous response to temperature change similar to that of free thermal strain. Several experiments showed rather clearly that the extra deformation caused by change in temperature distinctly occurs during the period of changing temperature, while the deformation under load after this period is just the basic creep (Illston and Sanders, 1973).

2.2.7. Decomposition of the total strain due to coupling effects

Decomposition of total strain of concrete due to combined thermal and mechanical loadings is a complicated issue, mainly because the definitions of the terminologies used in the literature are not consistent among researchers. In this report, we will restrict to the definitions and explanations given by Khoury (1985 a & 2002). When the temperature and sustained mechanical load are applied to the specimen simultaneously, the total thermo-mechanical strain will be observed. The decomposition of the total strain can be considered experimentally or numerically to estimate the effect of combined thermal and mechanical action or the influence of each strain component of the total strain. The total strain decomposes into the following components:

Total strain = FTS(Free Thermal Strain) LITS(Load Induced Thermal Strain) (2.1)

52

Figure 2.32. Strain of unsealed basalt concrete measured during heating at 1 C/min under three uniaxial compressive load levels (percent of strength prior to heating) excluding the initial elastic strain (Khoury, 2002)

Figure 2.33. Relative proportions of three load induced strains in uniaxial compression of concrete (Khoury, 2002)

53 Figs. 2.32 and 2.33 provide the conceptual argument of LITS. The definition for each terminology is as follows. FTS (Free thermal strain) - FTS is a function of temperature and time. It includes drying shrinkage (due to moisture loss) and expansive strains. Load induced thermal strain (LITS) - LITS is the strain that develops when concrete is heated for the first time under load. LITS is composed of transient creep, basic creep, and the elastic strain that occurs during the heating process. The term basic creep is used cautiously here, because this strain strictly occurs when concrete is loaded at constant temperature and after all internal reactions have been completed (Khoury 1985 b).

LITS = Transient Creep + Basic Creep + Elastic strain (during the heating) (2.2)

Transient creep is classified as transitional thermal creep and drying creep in case of unsealed specimen.

Transient Creep (unsealed specimen) = Transitional Thermal Creep (TTC) (sealed specimen) + Drying Creep
(2.3)

TTC (Transitional thermal creep) - TTC develops irrecoverably during, and for a few days following, first-time heating of sealed concrete under load. It appears in addition to the increase in elastic strain and basic creep (flow and delayed elastic) components with temperature. TTC is generally developed within a month from the

54 start of heating (Khoury, 1985 a). The strains developed under combined thermal and mechanical loading have not been considered systematically in the constitutive models for structural analysis.

2.2.8. The effects of steel and polymer fibers at elevated temperatures

Recent research has been performed by several researchers to improve the fire resistance of concrete. Polypropylene fibers and steel fibers are very useful materials for improving fire resistance of concrete. They are particularly useful to enhance the spalling resistance of HSC. Kodurs test results (2000) show that the addition of polypropylene fibers minimizes spalling in HSC members under fire conditions. One of the well-accepted theories is that by melting at a relatively low temperature of 170 C, the polypropylene fibers create channels for the steam pressure in concrete to escape, thus preventing internal pressurization causing spalling. The study showed that the amount of polypropylene fibers needed to minimize spalling is about 0.1 percent to 0.25 percent (by volume). On the other hand, the fibers increase the tensile strength of concrete (below the melting point of the fibers), which also provides higher resistance to the pore pressure. The presence of steel fibers increases the ultimate strain and concrete ductility according to the results of Lie and Kodur (1995 a). Cheng et al. (2004) carried out compression experiments to study the influence of various aggregates and steel fibers at elevated temperature in HSC. The test results were similar to that of NSC.

55

(a) with steel fibers

(b) without steel fibers

Figure 2.34. Failure mode with steel fibers (a) and without steel fibers (b) in HSC

Figure 2.35. The loading frame and the furnace (Fu-ping Cheng et al., 2004)

56

Figure 2.36. High strength concrete blocks after two-hour hydrocarbon fire test: (a) without polypropylene fibers; (b) with polypropylene fibers (Bilodeau et al., 1998)

Further research is being carried out to determine the optimum fiber content for different types of concrete. The effect of polypropylene fibers on spalling is illustrated in Fig. 2.36, which shows HSC concrete blocks after two hours of fire exposure.

2.3. Concrete Structures under High Temperatures

Local high temperature due to fire results in damage in concrete structural members such as beams, columns, walls, and slabs. The damage in concrete may be manifested as deterioration in strength or stiffness or as spalling. In the case of severe damage in key structural members, localized damage may lead to structural failure. In the past two decades, the research on the behavior of plain concrete under fire has

57 been conducted extensively at material level. However, the research on the behavior of reinforced concrete structures under fire (at structural level) has not been very active as compared to the research on plain concrete. Literature review indicates that the behavior of a structural member in a structure is different from that of a single member under the same fire conditions. As an example, the full scale test of an eight story steel structure was conducted at the Building Research Establishments Cardington Laboratory in England during 1995 and 1996 (Fig. 2.37).

Figure 2.37. A fire test of full-scale steel structure built in Cardington Laboratory (Grosshandler, 2002)

The test data demonstrated that the requirements for fire safety in structural design were overly conservative at the level of structural elements. The reason is that due to continuity, restraint conditions, and the interaction of members in a complete structure, an alternative load path is developed by the rest of the structure to bridge

58 over the members failed due to the local fire, thus enhancing the fire performance of the entire structure. Therefore, in principle, different local fire scenarios in a structural system should be considered in order to assess the realistic fire resistance of the structure. In practice, however, very limited experimental data on overall structural behavior in real fire conditions are available due to the complexity and cost of carrying out such experimental studies. Therefore, in the following sections we will focus on experimental results on behaviors of reinforced concrete structural members under high temperature. It is worthwhile to mention that the lack of experimental results on structural performance under high temperature has generated an interest in developing numerical models, which can be used to simulate different fire scenarios. Currently, there is a pressing need for developing reliable simulation models to characterize the behavior of concrete structures under high temperatures.

2.3.1. Factors affecting fire performance of concrete structures

The factors related to spalling-damage in concrete structures can be summarized as follows. Some of the factors were already introduced in Chapter 2.2.
Concrete strength: While it is difficult to specify the exact strength range,

based on the available information, concrete strengths higher than 55 MPa are more susceptible to spalling and may result in lower fire resistance.
Moisture Content: The moisture content, expressed in terms of relative

humidity, RH, influences the extent of spalling. Higher RH levels lead to greater regions of spalling. Fire-resistance tests on full scale HSC columns have shown that

59 significant spalling occurs when the RH is higher than 80 percent. The time required to attain an acceptable RH level (below 75 percent) in HSC structural members is longer than that required for NSC structural members because of the low permeability of HSC.
Type of Aggregate: For the two commonly used aggregates, carbonate

mineral aggregate (predominantly limestone) provides higher fire resistance and better spalling resistance than siliceous mineral aggregate (predominantly quartz). This is mainly because carbonate aggregate has a substantially higher heat capacity (specific heat), which is beneficial in preventing spalling.
Concrete Density: The extent of spalling of structural members made with

lightweight aggregate is much greater than concrete made of normal-weight aggregate. This is mainly because lightweight aggregate contains more free moisture, which creates higher vapor pressure under fire exposures.
Fire Intensity: High heating rates are critical for the occurrence of spalling.

Spalling of HSC is much more severe than NSC under the same heating rate.
Specimen Dimensions: The risk of explosive thermal spalling increases with

increasing specimen size. This is due to the fact that the size is directly related to the heat and moisture transport through the structural member. Larger members can store more energy. Therefore, careful consideration must be given to the size of structural members when evaluating spalling; fire tests are often conducted on small-scale specimens, which may provide misleading non-conservative results.
Lateral Reinforcement: The spacing and configuration of ties have

significant effects on the performance of HSC columns. Both closer tie spacing (at

60 0.75 times that required for NSC columns) and the bending of ties at 135 back into the core of the column, as illustrated in Fig. 2.38, enhance fire performance. The provision of cross ties also improves fire resistance. Fire tests on HSC columns, with additional confinement through cross ties and bending of ties back into the core of the column, have shown that spalling is significantly reduced (Kodur, 1998).

Figure 2.38. Tie configuration for reinforced concrete column: (a) conventional tie configuration (b) modified tie configuration (Kodur, 1999)

Load Intensity: A HSC structural member loaded in compression will spall

to a greater degree than an unloaded member because the mechanical stress adds to the thermal stresses, as shown in Fig. 2.22.

2.3.2. Fire scenarios

The idealized fire curves (standard fire curves) based on experience in real fires are used in fire testing, analysis and design. The standard fire curves differ from country to country and depend on the structural application. There are three main categories of fire curves: building, offshore/petrochemical and tunnel fires. Figs. 2.39

61 and 2.40 show the standard fire curves ISO 834 (or BS476) and ASTM E119 for buildings based on typical building fires, the hydrocarbon curve used in the offshore and petrochemical industries, and the RWS/RABT curves used in tunnels.

Figure 2.39. Standard fire curves for (ISO 834 (or BS476) and ASTM E119) based on a typical building fire (Grosshandler, 2002)

Figure 2.40. Standard fire scenarios for buildings (ISO 834 or BS 476), offshore and petrochemical industries (hydrocarbon), and tunnels (RWS, RABT) (Khoury, 2000)

62

The standard furnace curve represents a typical building fire in which the fuel source is usually wood, paper, fabric, etc. In ISO 834, the temperature increases from 20 C to 842 C after the first 30 min. (i.e., equivalent to an average heating rate of 27.4 C/min). This fire profile has a slow temperature rise up to 1000 C over a period of 120 min. This curve represents only one possible exposure condition at the growth and the fully developed fire stages, and does not include the final decay stage. By comparison, real fires may exhibit a slower or longer growth phase, and once they are established, the temperatures can be higher than the furnace temperature, though they are rarely sustained because they are subject to pronounced fluctuations. The standard temperaturetime curve, therefore, corresponds to a severe fire, but not to the most severe possible fire. In the 1970s, the Mobil Oil company investigated hydrocarbon fuel fires and developed temperaturetime profiles with a rapid temperature rise in the first 5 min. up to 900 C (i.e. 176 C/min) and a peak of 1100 C/min. This research laid the foundation for assessing the performance of fire protection materials in the offshore and petrochemical industries. In the Netherlands, the Ministry of Public Works, the Rijswaterstaat (RWS), and the TNO Centre for Fire Research established a fire curve for the evaluation of passive protection materials in tunnels. This RWS fire curve simulates the most severe hydrocarbon fire, rapidly exceeding 1200 C and peaking at 1350 C (melting temperature of concrete) after 60 min. and then falling gradually to 1200 C at 120 min., the end of the curve. The RWS fire curve was established on the basis of

63 Dutch experience in tunnel fires. However, the maximum temperatures attained in recent major fires did not reach RWS levels, e.g. Channel (1100 C), Great Belt (800 C), Mont Blanc (1000 C), Tauern (1000 C). The RWS fire curve, therefore, represents the most severe form of tunnel fire in terms of initial heating rates and maximum temperature. The RABT German fire curve in Fig. 2.40, with a descending branch, represents a less severe fire scenario in tunnels than the RWS fire curve, reaching a maximum temperature of 1200 C (melting point of some aggregates) sustained up to one hour before decaying to ambient conditions (Khoury, 2000).

2.3.3. Experimental studies on concrete beams

Sanjayan and Stocks (1991) conducted fire tests with two full-scale T-beams, one made of high-strength concrete with silica fume (105 MPa) and one made of normal strength concrete (27 MPa). For both specimens, the length was 2.5 m, the width of the flange was 1.2 m, the depth of the web from the top surface was 45 cm, and the width of the web was 25 cm. Different flange thicknesses of 20 cm and 15 cm were cast for both beam specimens on each side of the web. To study the effect of reinforcement cover, the cover of steel in the 20 cm flange was 7.5 cm, and for 15 cm flange the cover was 2.5 cm. In addition, the main bars in the web were staggered diagonally to obtain reinforcement covers of 2.5, 5.0, and 7.5 cm along the web. The beams were stored in the laboratory for three and half months prior to testing. Standard fire scenario ASTM E-119 was used and temperature profiles were obtained at 2.5, 5.0, and 7.5 cm from the bottom of the web with thermocouples. Weight loss

64 versus time was measured for both specimens while the test was in progress. Figs. 2.41 and 2.42 show the temperature profiles and weight loss versus time, respectively.

Figure 2.41. Temperature vs. Time at different locations (Sanjayan and Stocks, 1991)

Figure 2.42. Weight loss vs. Time in the test (Sanjayan and Stocks, 1991)

65 Moisture began to drop from several vertical cracks in both specimens at the average temperature of 695 C of the furnace (about 15 min. into the test). Explosive spalling occurred in the 20 cm flange of HSC specimen between 20 min and 40 min., but there was no spalling in the web, possibly because the distance for the moisture to escape was much shorter as compared to the flange due to exposure of three sides of the web for fire, and because wider flexural cracks existed in the web. However, in case of NSC, there was no spalling, even though the temperatures inside the NSC specimen were only slightly lower than in the HSC specimen. The range of spalling is indicated in Figs. 2.41 and 2.42. Hansen and Jensen (1995) conducted beam tests using three types of concretes in the Norwegian Fire Research Laboratory. The test specimens included reinforced and prestressed concrete beams having dimensions of 150 mm200 mm2850 mm. Three types of concrete were used, normal density concrete, lightweight aggregate concrete (Liapor aggregate), and lightweight aggregate concrete with fibers (Fibrin fiber types 1823). The concrete specimens (100 mm100 mm100 mm cubes) were designed to have a target 28-day cube strength of 75 MPa and 95 MPa. In addition, some beams were provided with a coating of Lightcem concrete (manufactured by LightCem A/S, Norway) for passive fire protection. Standard bars of quality K500TS according to the requirements of Norwegian Standard NS 3570 were used as reinforcement. The longitudinal reinforcement included 20 mm and 32 mm bars, and 8 mm stirrups. For prestressing, 26 mm, Dywidag bars, of the type St 835/1030 were used. A prestressing force of 300 kN was centrically applied at the ends of the beams, resulting in a mean prestress of approximately 10 MPa. Fire tests were performed in

66 an oil-heated furnace. The furnace has horizontal exposure openings with dimensions 2.5 m5.0 m and a depth of 1.5 m. The concrete beams were exposed to a hydrocarbon fire on three sides. The test procedure was in accordance with ISO 834. Thermocouples were installed on the longitudinal and shear reinforcement at two locations in each beam. The beams were exposed to the ISO 834 hydrocarbon fire for two hours, and the maximum oven temperature reached approximately 1100 C. The furnace temperature history and the period when spalling was observed are shown in Fig. 2.43.

Figure 2.43. ISO Hydrocarbon Fire Curve, Furnace Temperature, and Time Period when Spalling was Observed (Hansen and Jensen, 1995)

The test date obtained in this study showed that (a) severe spalling (exposed reinforcement) occurred in the high strength lightweight aggregate beams, while spalling without exposed reinforcement occurred in normal weight density concrete

67 beams; (b) shallow spalling or no spalling was observed for high strength lightweight concrete beams with fibers; and (c) no spalling was observed for the lightweight beams with fibers and the passive protective coating. Deep beams are structural elements loaded as beams in which a significant amount of the load is transferred to the supports by a compression thrust joining the load and the support reaction. As a result, the strain distribution is no longer linear along the depth of the beam, and the shear deformations become significant when compared to pure flexure. Floor slabs under horizontal load, short span beams carrying heavy loads, and transfer girders are examples of deep beams. Although a lot of studies have been conducted on deep beams, the structural behavior of deep beam is still not completely understood. In particular, the behavior of deep beams when exposed to high temperatures has not been studied very well.

Figure 2.44. Specimen geometry and thermal cycle (Felicetti and Gambarova, 1999)

Felicetti and Gambarova (1999) conducted high temperature deep beam tests with high strength concrete using siliceous aggregate (predominantly quartz), which is weaker than carbonate aggregate in terms of fire resistance. The purpose of the study was to find whether the deep beam fails in bending, shear-bending or shear-

68 compression in a high-temperature environment. Fig. 2.44 shows specimen geometry and thermal cycle used in their tests.

(a) Siliceous HSC

(b) Calcareous HSC

(c) Typical NSC

Figure 2.45. Strength of the three mix designs (Felicetti and Gambarova, 1999)

Compression tests and tension tests were conducted prior to the deep beam experiment. Fig. 2.45 shows compression and tensile strength values for three kinds
T 20 of concretes after a temperature cycle as the ratio f cT f c20 and f ct f ct . The residual

strength decreased with increasing target temperature in all tests. The test results of typical NSC were very interesting, in which the compressive strength at high temperature was higher than that measured after cooling down to room temperature (i.e. after a thermal cycle). Little difference between the hot and cold residual tensile strength was found under slow cooling condition of the residual specimens. However, no explanation was provided about these results. To determine the crack pattern in the beam tests, the back face of the beams was covered with Moir grid having 40 vertical lines/mm, and a reference grid (40.1 lines/mm) was superimposed. In this way, the interference between the distorted grid glued to the specimen and the reference grid produced a fringe pattern with a

69 sensitivity of 1/80 mm. Cracks appear as discontinuities in the fringe patterns. Fig. 2.46 shows fringe pattern and cracking of specimen at T=20 C, and Fig. 2.47 shows the extent of cracking at failure by mechanical load after one cycle heating up to 500 C. Fig. 2.48 shows load-displacement curves at the bottom face of mid span and the crack patterns.

Figure 2.46. Fringe pattern and cracking of specimen at T=20 C (Felicetti and Gambarova, 1999)

Figure 2.47. Cracking at failure by mechanical load after one cycle heating up to 500 C (Felicetti and Gambarova, 1999)

70

Figure 2.48. Load-displacement curves at bottom face of mid span and crack patterns (Felicetti and Gambarova, 1999)

The load-displacement diagrams show a marked decay of the mechanical behavior after a high temperature cycle. However, up to the 400 C cycle the ultimate load capacity (peak load) was only marginally affected, while the stiffness was more temperature-sensitive. Above 400 C the load capacity decreased dramatically, and the overall behavior became very ductile. Such occurrence is certainly not ascribable to the reinforcement, but to the loss of compressive strength in the concrete. In their test results, the collapse mode was mostly in the form of bending (20 C and 250 C) or shear-bending (400 C). Under the heating condition 500 C the beam exhibited a completely different behavior, since the mechanical decay of the mortar was accompanied by chemo-physical transformations in the highly siliceous aggregates, whose color and surface turned from white to red, and from glossy to opaque, with many particles split, owing to the expulsion of the bound water. In the end, failure was triggered off by the lack of compressive strength in one of the inclined struts. The experimental evidence showed that different ultimate behaviors may occur, depending on the severity of the thermal cycle. Three models were considered

71 in their paper (see Fig. 2.49). The first refers to the bending failure which assumes linearity of the strain distribution in the mid-span section (which is consistent with the actual value of the ratio h/L ); the second refers to the shear-compression failure, and is based on a strut-and-tie resistant system; and the third predicts the shearbending failure, assuming an inclined shear plane (beam-end model).

Figure 2.49. (a) Flexural model; (b) strut-and-tie model; (c) beam-end model

Figure 2.50. Comparison of failure models (Felicetti and Gambarova, 1999)

72 The comparison between the test results and the three models is shown in Fig. 2.50. The temperature level T = 400 C forms a threshold condition. Below this value bending failure (flexural model) is the only failure mode, while above this value other failure modes become more or equally probable. Between 400 C and 450 C bending (flexural failure mode) and shear-bending modes (beam-end mode) seem to be equally probable. The high reduction of the compressive strength above 300 C results in a downward trend of the ultimate load capacity of the strut-and-tie mechanism. The shear-compression mode (strut-and-tie model) becomes as probable as the other two failure modes after a cycle at 450 C. Above 450 C, the deep beam tends to fail definitely in shear-compression.

2.3.4. Experimental studies on concrete columns

Diederichs et al. (1995) reported that the use of fibers helps to reduce spalling in HSC columns and suggested that further studies be conducted on the effects of fiber contents. The document ENV 1992-1-2 (Structural fire design developed by European Commission for Standardization (CEN)) is essentially based on tabulated data containing the dimensions of the cross sections and values for the concrete cover. Dotreppe et al. (1996) conducted experimental studies on the determination of the main parameters affecting the behavior of reinforced concrete columns under fire conditions. The parameters used were the load level, the dimensions of the cross section, the length of columns, the main reinforcement, the concrete cover and the

73 eccentricity of the axial load. The concrete mix design used in the test program is summarized in Table 2.5.

Table 2.5. Mix design (Dotreppe et al., 1996) Gravel 4/14 Coarse sand Cement P40 Water 1100 kg 800 kg 280 kg 185 kg

The fire resistance tests were performed according to the Belgian Standard NBN 713.020 and the columns were exposed to fire on four sides. The support conditions of the columns were hinges at both ends. The dimensions of the cross section were: 200 mm300 mm, 300 mm300 mm, and 400 mm400 mm. The

column lengths were 3.95 m and 2.10 m. Concrete covers of 25 mm, 35 mm, and 40 mm were used. Eccentricities of (-20 mm, 20 mm), (20 mm, 20 mm), and (0 mm, 0 mm) were used with respect to the central axis of the 300 mm300 mm cross section. Experiments on 3.95 m high columns were performed at the University of Ghent. Fig. 2.51 shows the furnace used for the tests on the RC columns at the University of Ghent. The 2.10 m high column experiments were performed at the University of Lige. Fig. 2.52 shows the furnace used for the tests on columns at the University of Lige. Longitudinal reinforcing bars of 12, 16, and 25 mm in diameter were used (see Dotreppe et al., 1996 for detailed information). Although the type and arrangement of stirrups was not a parameter, buckling of some individual longitudinal reinforcements may occur between two stirrups at column failure.

74

Figure 2.51. Furnace used for tests on columns at the University of Ghent: (a) Overall view (b) Detail of top hinge and load cell (2000 kN) (c) Detail of bottom hinge (measurements in millimeters) (Dotreppe at el., 1996)

Figure 2.52. Furnace used for tests on columns at the University of Lige: (1) Lower transverse beam (2) Jack with double effect (3.1) Lateral support (3.2) Supports perpendicular to frame plate (4) Load cell (5) Upper transverse beam (6) Support for upper transverse beam (7) Furnace (8) Crossing column (Dotreppe at el., 1996)

75

Therefore, decreasing the spacing between stirrups might improve the behavior of the column under fire conditions, however theoretical and experimental research should be performed in order to quantify this effect. A summary of the test results follows: - In all problems involving fire resistance, the load level (column stress) is the most important factor. The fire resistance decreases when the load increases. - The dimensions of the cross section influence the fire resistance. The fire resistance of 200 mm300 mm sections was between 1 hour and 2 hour, while in most of the 300 mm300 mm sections the fire resistance was greater than 2 hours. However the applied load must be limited and reinforcement with large diameters rebars should be avoided. - In the case of 400 mm400 mm columns, the fire resistance time was appreciably shorter than expected. For one of the columns, the fire resistance time was shorter than 1 hour, and for the other, the fire resistance time was between 1 hour and 2 hours. These inconsistent results may be partly explained by the existence of reinforcement with large diameters in one case, and by the use of a section with large dimensions and small concrete cover in both cases. - The increase of column length had a negative influence on column failure at normal as well as at high temperature because of geometrically nonlinear effects. - In principle, the increase of the concrete cover should result in an increase of the fire resistance or of the admissible load level, since the temperature in the main reinforcement increases less rapidly when the column begins to deflect laterally. The

76 test results showed that the increase of concrete cover had a positive effect on the fire resistance. - The use of eight 16 mm diameter reinforcement instead of four 25 mm diameter reinforcement showed a positive effect on the fire resistance in terms of strength. But, additional tests on columns involving reinforcing with large diameter bars are still required to clarify the effect of the bar diameters of longitudinal reinforcement. - Test results of eccentrically loaded columns (-20 mm, 20 mm) and (0 mm, 0 mm) showed approximately the same failure results. On the contrary, the use of eccentricities (20 mm, 20 mm) exhibited a decrease of the load capacity levels.

(a) NSC column

(b) HSC column

Figure 2.53. NSC and HSC column after ASTM E119 fire tests (Kodur, 1999)

77 Kodur (1999) conducted an experimental study to compare the fire resistance of NSC and HSC columns. He also proposed guidelines for improving the fire resistance of HSC structural members. Tests were performed under both laboratory and actual fire conditions. Fig. 2.53 shows the conditions of HSC and NSC columns after concluding the fire resistance tests. In the HSC column, the reinforcement (both longitudinal and transverse) is completely exposed and there is significant spalling. However, there was no spalling in the NSC column. Since spalling occurs during the initial stage of fires in HSC columns, it may pose a risk to evacuating occupants and firefighters. As discussed in Chapter 2.2, spalling is attributed to the building up of pore pressure during heating. HSC is believed to be more susceptible to this pressure build-up because of its low permeability.

Figure 2.54. Temperature distribution at various depths in NSC and HSC columns (Kodur, 1999)

78

Figure 2.55. Axial deformation histories of NSC and HSC columns during fire exposure (Kodur, 1999)

Fig. 2.55 shows that the axial deformation of the HSC column is significantly lower than that of the NSC column. This can be explained in part by the lower thermal expansion of HSC and the slower rise in temperature in the HSC column during the initial stages of fire exposure due to the compactness of HSC. As the temperature is continuously rising, the steel reinforcement in the NSC column and the HSC column gradually reaches the yield capacity and the RC column contracts. When this happens, the concrete carries a progressively increasing portion of the axial load. The strength of the concrete also decreases with time and, ultimately, when the column can no longer support the load, failure occurs. At this stage, the column behavior depends primarily on the strength of concrete. There is significant contraction in the NSC column leading to gradual (ductile) failure. The lower contraction in the HSC column can be attributed to the fact that at elevated

79 temperatures HSC becomes brittle and loses strength faster than NSC. For the NSC columns, the fire resistance time to failure was approximately 366 min, while for HSC column it was 225 min (Kodur, 1999).

2.3.5. Experimental studies on concrete slabs

Di Prisco et al (2003) conducted tests to study the thermal and mechanical behavior of plain concrete slabs and steel fibers reinforcement concrete slabs, SFRC. The concrete were loaded mechanically and then exposed to a standard fire up to failure.

Figure 2.56. Position of thermo couples and load set-up (di Prisco et al., 2003)

80 The size of the slab specimens was 1800 mm 600 mm 60 mm. The purpose of these tests were (a) to find thermal parameters of SFRC as compared to the plain concrete, and (b) to obtain data on the structural behavior under fire which can be used as a reference to validate computational models. Fig. 2.56 shows the furnace, the position of thermo couples and the load set-up of the slab experiments. Fig. 2.57 compares the standard time-temperature curve and the actual heating curves in the three furnace compartments. In the experiments, the actual mean temperature-time curve was satisfactorily close to the standard heating curve (ISO 834). The temperature dips in Test 2 and Test 3 occurred by a temporary flame-out of the burners due to debris falling from the specimens.

Figure 2.57. Comparison between the standard time-temperature curve and the actual heating curves (di Prisco et al, 2003)

Fig. 2.58 shows on the left the temperature distribution at different depth levels inside Test 1-SFRC slab (unloaded). At the right it illustrates the vertical displacement of the SFRC slab as compared to the plain-concrete slab under timetemperature conditions (Test 2 and Test 3) and application of a bending moment that

81 corresponds to some portion (11, 24, and 40 percent, di Prisco et al., 2003) of the ultimate bending moment at the reference temperature. The experiments were aimed at quantifying the time required to failure for different levels of mechanical loading. The temperature distribution curves at different depths of the SFRC slab show parabolic trends, which are similar to the actual heating curve (Fig. 2.57, Test 1). The SFRC slab was more ductile than the plain concrete slab due to the addition of steel fibers. The performance of the test slabs during the temporary flame-out of the burners was not jeopardized, as indicated in Fig. 2.58 (under load).

Figure 2.58. Temperature distribution according to depth inside SFRC slab and vertical displacement of SFRC and plain concrete slabs (di Prisco et al, 2003)

Fig. 2.59 shows the mean value of the load-displacement curves of the SFRC slabs (using low carbon fibers) in 4-point bending in the virgin state and after exposure to high temperature (T = 600 C) but tested after cooling. The temperature effects under hot and residual test conditions were very large as compared to the flexural test performed at room temperature. The fire resistance after cooling was slightly higher than that at high temperature. However, the general behavior of the

82 SFRC slab did not exhibit a large difference between residual and hot testing as shown in Fig. 2.59.

Figure 2.59. Virgin, hot and residual load-deflection of slabs (di Prisco et al, 2003)

Foster et al. (2004) conducted an investigation on the influence of thermal curvature on the failure mechanisms of rectangular slabs. The loading frame and test set-up are shown in Fig. 2.60. Fig. 2.61 shows the heating elements and the supporting frame. To satisfy similar boundary conditions to slabs used in the eight story steel structure tested at the Cardington Laboratory (see Fig. 2.37), the slab was placed on a supporting frame, providing vertical support around the perimeter. The four corners of the slab were loosely clamped to restrain the vertical upward movement without restraining the horizontal movement at the support edges. The applied load remained vertical with the aid of ball-joints which allowed the loading system to rotate as the slab deflected.

83

Figure 2.60. Loading frame and test set-up (Foster et al., 2004)

Figure 2.61. Supporting frame and heating elements (Foster et al., 2004)

Fig. 2.62 shows the locations of the applied loads and displacements gages (horizontal and vertical displacement). The slabs were reinforced with smooth or deformed (ribbed) steel wire meshes which were arranged isotropically. The deformed wire was made by indenting the smooth wire using a purpose-built machine. Fourteen slabs were used for testing and one of them was used to obtain a temperature profile without loading.

84

Figure 2.62. Locations of loads and displacements gages (Foster et al., 2004)

Fig 2.63 shows the temperature history profile at different positions at each of the three levels through the thickness of the slab tested without mechanical loading.

Figure 2.63. Temperature profiles (Foster et al., 2004)

Thirteen slabs were loaded before the heating elements were switched on. The formation of yield lines was depending on the applied load level. After the initial slab

85 deflected into double curvature a crack was generated across its short span through the full-depth. Thereafter a less distinctive yield line mechanism developed at high temperature which resulted into significant thermal bowing. All slabs behaved similarly.

(a) Yield line mechanism

(b) Topside

(c) Bottom

Figure 2.64. Yield line mechanism and crack patterns (Foster et al., 2004)

Approximately 15 min. after the furnace was switched on, diagonal cracks occurred across the corners as the slab deflected into double curvature. After 20 min., a single transverse crack could be seen forming on the topside of the slab in its central

86 region. Over time this crack developed outwards in the short-span direction towards the long edges of the slab (Fig. 2.64 (b)). Most of the slabs developed crack patterns resembling a yield line mechanism towards the end of the test, usually occurring after 2 hours. The crack patterns resemble those of similar slabs tested at ambient temperature. The mid-span deflections for the tested specimens are shown in Fig. 2.65. The rate of deflection of the slab reinforced with smooth wires (Test 13) was less than the equivalent slab using deformed wires (Test 12). The reinforcing bars across the large tensile cracks were exposed to high temperature early on. These results imply that slabs reinforced with smooth wire meshes perform better than those reinforced with deformed wire meshes.

Figure 2.65. Mid-span deflections for the tested slabs (Foster et al., 2004)

87

CHAPTER 3

3.

AN EXPERIMENTAL STUDY FOR THERMO-HYGRO COUPLING EFFECT OF CONCRETE AT ELEVATED TEMPERATURE

3.1. Introduction

At elevated temperature, the thermal strain of concrete is related to changes of both temperature and moisture. The measurements of thermal expansion of solid materials in ASTM [ASTM E 831-03, C 531-00, and E 228-95] are obtained by measurement of the displacement change of the material in accordance with temperature change from initial temperature up to a target temperature without moisture control. Usually, the thermal strain of concrete (CTS), is obtained by measurement of the displacement change without moisture control. It is not easy to experimentally measure the temperature effect and moisture effect on strain separately over time. In this experimental study, CTS, the strain caused by temperature increase under constant humidity (PTS), and the strain caused by moisture change under constant temperature (PHS) are measured continually and simultaneously over time. The maximum temperature in the chamber, which can control both humidity and temperature simultaneously, is 80 C. Since high heating

88 rates will lead to large specimen temperature gradients, a low heating of rate 0.1 C/min is used for the test. The objectives of this experimental study can be explained as follows. The first objective is to measure the strain as the temperature and humidity changes continually and simultaneously over time. It appears that there has been no experimental study adopting such a testing method. The second objective is to examine the difference between CTS and PTS experimentally. The third objective is to identify the thermo-hygro coupling effect based on test data and analysis.

3.2. Experimental details

3.2.1. Material and mixture portion

Only one aggregate (granite) is used for this test. With respect to density aggregates are classified as light weight (less than 1120 kg/m3), normal weight (15201680 kg/m3) and heavy weight aggregates (larger than 2080 kg/m3). Most of normal weight aggregates such as limestone and granite are classified as natural mineral aggregate. For this experiment, granite is chosen over limestone due to the advantages of being more durable and the fact that it readily bonds to concrete mix. Fig. 3.1 shows the granite sand and gravel used in the research. Table 3.1 shows typical proportions of materials for producing low, moderate, and high-strength concretes using normal-weight aggregates. Among the three mix designs, the moderate-strength mix design is used for the present test. However, the water/cement (w/c) ratio used

89 for this experiment is different (0.71 by weight) from the table because the workability could not be obtained from the mix design; this is due to dry aggregates and the non-negligible contents of very fine size aggregates in crushed granite sand.

Figure 3.1. Granite sand and gravel

Table 3.1. Typical proportions of materials for normal weight concretes

To calculate the real w/c ratio considering moisture contents of the aggregates, the oven dry (OD) and saturated surface dry (SSD) moisture contents are measured. The test for OD moisture content follows test method B in ASTM D 2216. The moisture contents for OD and SSD are calculated by Eqs. (3.1) and (3.2).

90

MC (OD) = [( M cms M cds ) /( M cds M c )] 100 (%) MC ( SSD) = [( M cms M css ) /( M css M c )] 100 (%)

(3.1) (3.2)

Where MC (OD) is the moisture content based on oven dry (%), MC ( SSD) is the moisture content based on saturated surface dry (%), M cms is the mass of container and moist aggregates (g), M css is the mass of container and saturated surface dry aggregates (g), M cds is the mass of container and oven dry aggregates (g), and M c is the mass of container (g).

MC ( SSD) and MC (OD) obtained from above equations are -0.36 % and
0.42 % respectively. The negative value of MC ( SSD) means that pores in the concrete are partially filled with water. Absorption capacity (AB) for the aggregates is then calculated by Eq. (3.3). Absorption capacity is a measure of the mount of water in pores.

AB = (WSSD WOD ) 100 (%) / WOD

(3.3)

In which, WSSD is the unit weight of aggregates in saturated surface dry and

WOD is the unit weight of aggregates in oven dry. The absorption capacity of
aggregates is about 0.78 %. By subtracting the water absorption capacity of the aggregates from original water content, the finalized real w/c ratio is 0.67 by weight. The maximum size of coarse aggregates is 3/8 inch.

91

3.2.2. Specimen preparation and test equipments

Fig. 3.2 shows the relative humidity variation over time measured at the center of a 24 inch cylinder and a 48 inch cylinder under standard room conditions.

100 90 80 70 RH (%) 60 50 40 30 20 10 0 0 20 40 60 Time (Days) 80 100

2*4 cyinderical specimen 4*8 cylinderical specimen

Figure 3.2. Relative humidity of 24 in. and 48 in. cylinder vs. Time

Using a 24 inch cylindrical specimen reduces the testing time because of the rapid humidity reduction ratio compared with 48 inch cylindrical specimen, as shown in Fig. 3.2. Four specimens are used for each test. Two specimens are used to measure relative humidity and temperature from a sensor embedded at center of each specimen. Two other specimens are used to measure the strain from two strain gages attached on surface of each specimen. The model SHT 75 sensor and EK-H3 data logger produced by SENSIRION as shown in Fig. 3.3 are used to measure temperature and relative humidity. The

92 sensor can measure both temperature and relative humidity simultaneously. The capacities of the sensor recommended by the manufacturer are 0-100 % for relative humidity and -40-120 C for temperature. The strain gage (Model N2A-06-20CBW120) by VISHAY and the data logger (Model CR10X) by CAMBELL SCIENTIFIC are used to measure strain.

Figure 3.3. SHT 75 sensor and EK-H3 data logger

In the specimens with a sensor, a plastic tube is used to keep the sensor settled in center of the specimen because the simple cable connection is too loose for the purpose of this experiment. The cable is inserted through the plastic tube until the sensor connected to the cable reaches the other side, then the gap between the cable and tube on end sides of the tube are sealed using sealant (with temperature resistance up to 200 C) to prevent moisture and temperature loss through the tube. The sensor is wrapped with GORE-TEX fabric to protect it from being damaged from concrete paste. Fig. 3.4 shows the plastic tube built with the cable and sensor, and the 24 inch cylindrical specimen embedded with a sensor. The specimens are cured in a curing room (Temp. 24 C and RH 92 %) for 28 days.

93

Figure 3.4. Plastic tube built with the cable and sensor and 24 inch cylindrical specimen embedded a sensor at center

Figure 3.5 Equipments and test set up

The RUSSELLS environmental chamber (Model RD-5-1) is used to control relative humidity and temperature. The temperature and humidity control range of the chamber is from -30 C to 177 C and 0 % to 95 % respectively. Fig. 3.5 shows the equipments and test set up.

94
3.2.3. Thermal compensation function and verification for test method

The data obtained from the strain gage require thermal compensation because the grid alloy used on the strain gage also changes with temperature. The strain gage used is a resistance strain gage. Since the gage grid is made from a strain-sensitive alloy, the resistance change of the strain gage is also proportional to the thermally induced strain. The strain function with thermal compensation for concrete is obtained from the following derivation. When the thermal expansion coefficient of the gage grid is different from that of a specimen used for the test, the net resistance change of the gage for temperature can be expressed as follows:

R = [ G +FG ( S G )] T Ro

(3.4)

In which, R Ro is the net resistance change of gage, G is the thermal coefficient of resistivity of grid material, FG = FR [1 + c (T 24C )] (gage factor as calculated in an uniaxial stress state), c = 1.2 106 / C (temperature coefficient of gage factor), FR is the gage factor at room temperature 24C (given as 2.1 by manufacturer), S G is the difference in thermal expansion coefficients between specimen and grid, and T is the temperature change from arbitrary initial reference temperature. Thus, thermal output in strain units is expressed by Eq. (3.5).

95

(T ) G / S =

R / Ro [ G +FG ( S G )] T = FI FI

(3.5)

In which, (T ) G / S is the thermal output for grid gage G on specimen material S and FI is the instrument gage factor setting employed in recording thermal output data (2.0 for all alloy gages). From Eq. (3.5), the thermal outputs of the concrete and 1018 steel are:

(T ) G / CONCRETE =

[ G +FG ( CONCRETE G )] T (for concrete) FI

(3.6)

(T ) G / 1018 STEEL =

[ G +FG ( 1018 STEEL G )] T FI

(for 1018 steel) (3.7)

By subtracting thermal output of the 1018 steel from that of the concrete, the thermal strain function of the concrete with thermal compensation is expressed as Eq. (3.8).

(T ) CONCRETE =

FG [ (T ) G / CONCRETE (T ) G / 1018 STEEL ] + 1018 STEEL T FI

(3.8)

The coefficient of thermal expansion of 1018 steel is 12 10 6 / C and

(T ) G / 1018 STEEL is given from the thermal output data sheet of the strain gage tested on
1018 steel provided from the manufacturer.

96 To verify the thermal compensation function and test method, the thermal expansion of aluminum is tested with the same strain gage (model N2A-06-20CBW120) as the concrete test. The range of the coefficient of thermal expansion for aluminum is from 21 10 6 to 25 10 6 / C .

100 90 80 RH (%) & Temp (C) 70 60 50 40 30 20 10 0 0 50 100 150 200 250 Time (hrs)
Temp (Chamber) RH (Chamber) Aluminum Strain

1600 1400 1200

Strain (x10-6)

1000 800 600 400 200 0

Figure 3.6. Temp, RH and Strain vs. Time (Aluminum)

Fig. 3.6 shows the result of the aluminum thermal expansion test. The aluminum sample is equilibrated to ambient temperature inside the chamber (about 28 C) and then the temperature inside the chamber is increased up to 80 C without moisture control. Fig. 3.6 shows the change of the ambient temperature, relative humidity inside the chamber, and aluminum thermal strain over time. The coefficient of thermal expansion of the aluminum, from the thermal compensation function, is found to be 22.5 10 6 / C which falls in the known range given above.

97

3.3. Test results and discussion

3.3.1. Conventional Thermal Strain (CTS) and Pure Hygro Strain (PHS)

As mentioned earlier, temperature and humidity histories at the center of the specimens and inside the chamber are measured over time. The strains of the concrete specimens are measured from the surfaces of the concretes because setting-up strain gages in the center of specimens is not possible. The strain of the specimens are measured under temperature control (heating rate 0.1 C/min and maximum temperature 80 C) without humidity control inside the chamber. After 28 days of curing, the relative humidity of the specimens in the curing room is about 92 % at 24 C. The specimens are taken out of the curing room and left in the laboratory until the relative humidity measured at the center of the specimens reaches the target relative humidity of 70 % at 24 C. Then, the specimens are placed in the chamber and kept for 4-5 days allowing them to reach thermal equilibrium at an ambient temperature of about 28 C. After the equilibrium state is reached, the temperature inside the chamber is increased to the target temperature of 80 C at a rate of 0.1 C/min without humidity control. Fig. 3.7 shows the histories of relative humidity and temperature inside the chamber. Relative humidity (RH) is expressed as the ratio of the partial vapor pressure to the saturated vapor pressure [see Eq. (3.9)].

98
100 90 80 RH (%) & Temp (C) 70 60 50 40 30 20 10 0 0 50 100 Time (hrs) 150 200 250
Temp (Chamber) RH (Chamber)

Figure 3.7. Temperature and RH inside the chamber vs. Time (No humidity control)

RH =

v v , sat

Pv wv P = RT = v Pv , sat wv Pv , sat RT

(3.9)

In which, v is the vapor density, v , sat is the saturated vapor density, Pv is the partial vapor pressure, Pv ,sat is the saturated vapor pressure, wv is the molar mass (molecular mass) of water vapor, R is the universal gas constant, and T is the temperature in Kelvin. The saturated vapor pressure is a temperature function that is the maximum amount of humidity air can hold at the specific temperature. The saturated vapor pressure increases exponentially as the temperature increases in a closed system, therefore the increase of the temperature generally leads to the decrease in relative humidity as shown in Fig. 3.7.

99

100 90 80 RH (%) & Temp (C) 70 60 50 40 30 20 10 0 0 50 100 Time (hrs) 150 200 250
Temp (Specimen) RH (Specimen) Strain_average

1200 1000 800 Strain (x10-6) 600 400 200 0 -200 -400

Figure 3.8. Temperature, RH and strain history vs. Time (No humidity control)

100 90 80 RH (%) & Temp (C) 70 60 50 40 30 20 10 0 0 50 100 Time (hrs) 150 200 250
Range A Range C

1200
Temp (Chamber) RH (Chamber) Temp (Specimen) RH (Specimen) Strain_average
Range B

1000 800 Strain (x10-6) 600 400 200 0 -200 -400

Figure 3.9. RH, temperature, and strain history vs. time (No humidity control)

100
5
RH (Chamber) RH (Specimen)

4 RH (%) & Temp (C)

0 170 180 190 Time (hrs) 200 210

Figure 3.10. Part C magnified in Fig. 3.9.

Fig. 3.8 shows the histories of the internal temperature and internal relative humidity at the center of the specimen, and the strain on the surface of the specimen. Comparing Fig. 3.8 with Fig. 3.7, one can see that when the temperature increases, the chamber RH decreases in Fig. 3.7, while the internal RH of the specimen increases. Fig 3.9 is the overlap plot of Fig. 3.7 and Fig. 3.8, which shows the two opposite RH variations more clearly. The heating period in Fig. 3.9 is called Range A, in which the internal relative humidity of the specimen increases with temperature increase. After Range A, it is Range B in which the temperature is kept constant and the internal RH decreases. Range C follows Range B, which is magnified and shown in Fig. 3.10. In Range C, when the temperature drops the internal relative humidity in the concrete drops, too. These three ranges in Fig. 3.9 need more detailed explanation.
Range B shows the diffusion process of the internal RH under constant

temperature. The diffusion of moisture is driven by the moisture gradient (i.e. higher internal RH in the concrete and lower external RH in the chamber). The measured

101 strain in Range B is the Pure Hygro Strain (the so-called true shrinkage strain without the effect of thermal expansion). In Range A and Range C, the internal relative humidity in concrete is changed in proportion to the temperature variation. The experimental study for hardened cement paste by Grasley and Lange (2004) shows the same trend, that the internal relative humidity in hardened cement paste increases as temperature increases. The variation of internal RH in Range A and Range C cannot be explained simply based on Eq. (3.9), which is valid in a large closed system and not suitable for porous material with very small pores. To this end, the relationship between RH and temperature in porous medium needs to be introduced. For temperatures below the critical point of water (374.15 C), concrete is distinguished as saturated and nonsaturated concrete. But, for higher temperature there is no such distinction because the liquid phase does not exist (Baant, 1996). In a partially saturated pore system, if it is assumed that a local thermodynamic equilibrium always exists between the phases of pore water (vapor, liquid) within a very small element of concrete, the static force equilibrium based on a capillary tube model is expressed with Eq. (3.10) [see APPENDIX A for detailed explanation].

= ( Pa Pw ) =

2Ts cos r

(3.10)

In which, ( Pa Pw ) , called suction, is the pressure difference between air and water in a pore, Ts is the surface tension of the pore fluid, r is the average radius of meniscus, and is the contact angle between air and water.

102 While mechanical equilibrium is expressed by Eq. (3.10), physicochemical equilibrium between the vapor and liquid is satisfied according to the Kelvin equation, Eq. (3.11) [see APPENDIX A].

2Ts cos RT = ln( RH ) r vw

(3.11)

In which R is the universal gas constant and vw is the molar volume of water. Thus, the relative humidity in a partially saturated porous medium from Eq. (3.10) and (3.11) is expressed by Eq. (3.12).

( P - P )v 2T cos vw RH = exp - a w w = exp - s RT r RT

(3.12)

Eq. (3.12) has been widely used to describe the relationship of temperature and relative humidity in partially saturated pore systems such as partially saturated soil (Lu and Likos, 2004), cement paste, and concrete [Power and Brownyard (1947), Baant (1996), Grasley and Lange (2004)]. From Eq. (3.12), it can be seen that under constant ( Pa Pw ) (or Ts , , and r ), when temperature increases, the RH in concrete increases. However, ( Pa Pw ) (or Ts , , and r ) depends on temperature. When temperature increases, water expands (swelling pressure), which means that ( Pa Pw ) decreases, the surface tension ( Ts ) of water decreases, the contact angle ( ) increase, cos decreases, and the radius of the meniscus curvature increases

103 (Power and Brownyard, 1947). All these changes in Ts , , and r lead to an increase in internal RH in concrete, which is the same as the direct effect of increasing T in Eq. (12). Fig. 3.11 shows a schematic for dilatation of solid microstructure induced by a decrease of capillary tension with a temperature increase.

RH= 50% Swelling Pressure Water

Capillary tension

RH= 80%

Swelling Pressure

Capillary tension

Water

Soild

Soild Increasing temperature

Soild

Soild

Figure 3.11. Dilatation of solid microstructure induced by decrease of capillary tension with temperature increase

Figure 3.12. Length change of Portland cement paste specimens at elevated temperature: (a) Philleo (1958); (b) Harada et al. (1972); (c) Cruz and Gillen (1980); Crowley (1956)

104

Figure 3.13. Linear thermal expansion of various rocks at elevated temperature: (a) sand stone; (b) limestone; (c) granite; (d) anorthosite; (e) basalt; (f) limestone; (g) sandstone; (h) pumice (Soles and Gellers, 1964)

Under the same concept Dettling (1964) attributed the initial expansion (see Fig. 3.13) of hardened cement paste, up to about 150 C, to kinetic molecular movements in the paste plus an increase of swelling pressures. When we consider the relationship between temperature and strain for aggregates and hardened cement paste as components of concrete, it is clear that the internal relative humidity (RH) of concrete increases as temperature increases. Fig. 3.12 and 3.13 show the length change of hardened cement pastes and aggregates at elevated temperature, respectively. Almost all of the aggregates expand as the temperature increases up to about 800 C. While initially the hardened cement pastes expand when heated to about 150 C, beyond 150 C the hardened cement pastes shrinks. This means that there is no shrinkage of concrete by dehydration (a

105 decrease of relative humidity) during heating up to about 150 C. It is also expected that the shrinkage effect due to a decrease in internal relative humidity (dehydration) of cement is contained in the overall expansion of concrete beyond 150 C.

600
CTS CTS_Linear

500

400 Strain (x10-6)

300

200

100

0 20 30 40 50 Temperature (C) 60 70 80

Figure 3.14. Conventional thermal strain by temperature increase from 28 C to 70 C

Fig. 3.14 shows CTS at elevated temperatures from 28-70 C from Range A in Fig. 3.9. The CTS curve in Fig. 3.14 is not exactly linear because of the moisture effect. Pure Hygro Strain (PHS), which is the moisture effect under constant temperature measured in this study, is nonlinear. For temperatures ranging of 2870 C, the linear function for the CTS is expressed as Eq. (3.13). T is the temperature in Celsius.

CTS = (12.85 T 359.8) 106

for 28 C T 70 C

(3.13)

106 From Range B in Fig. 3.9, Pure Hygro Strain (PHS) under constant temperature 77.5 C is plotted in terms of internal relative humidity of concrete in Fig. 3.15. In Range B, the fluctuation range of the internal temperature of concrete is 77.5 C 0.5 . The PHS function in terms of RH (%) from curve fitting is expressed as Eq. (3.14). For RH 5-84 % at constant temperature 77.5C, the PHS is:

PHS = 3 1011 ( RH ) 4 + 7 109 ( RH )3 5.463 107 ( RH ) 2

+ 1.953 107 ( RH ) 7.614 107

(3.14)

600
PHS_constant temperature 77.5C

500

400 Strain (x10-6)

300

200

100

0 0 10 20 30 40 50 60 70 80 90 RH (%)

Figure 3.15. Pure-Hygro Strain (PHS) under constant temperature 77.5 C 0.5

Theoretically, PHS should be linear with respect to humidity under constant temperature. An interesting result is that the PHS according to humidity increase is nonlinear. It is difficult to explain the trend clearly. The measurement for the strain

107 caused by humidity loss under constant room temperature shows the same result, that the strain caused by humidity loss is nonlinear.

100 90 80 RH (%) & Temp (C) 70 60 50 40 30 20 10 0 0 100 200 300 400 500 600 700 Time (hrs)
Temp (Room) RH (Room) Temp (Specimen) RH (Specimen) Strain_average

1400 1200 1000 800 600 400 200 0 -200 -400 Strain (x10-6)

Figure 3.16. RH, temperature, and strain history vs. time

0 Strain_average -50 Trend line

-100 Strain (x10-6)

-150

-200

-250

-300 30 35 40 45 50 55 RH (%) 60 65 70 75 80

Figure 3.17. Shrinkage strain according to internal RH decrease of concrete under constant temperature 24.5 C

108 Fig. 3.16 shows the strain due to humidity loss under constant room temperature 24.5 C. The strain versus internal relative humidity in concrete is plotted in Fig. 3.17. The strain, which is measured on the surface of the specimen, shows fluctuation relative to violent fluctuation of the relative humidity curve of the laboratory room. From the trend line in Fig. 3.17, it is shown that the strain due to humidity loss under constant room temperature 24.5 C is also nonlinear. It is likely linked to complicated physiochemical reactions in concrete with humidity change.

3.3.2. Pure Thermal Strain (PTS)

PTS is measured using dried specimens. Prior to the test the specimens are dried in an oven at a constant temperature of 102 C for 3 days. The specimens are then left in standard room conditions to come to equilibrium for 1 day, at a temperature of about 24 C. The test results are presented in Fig. 3.18 through 3.20 in the same format as those of the tests done under no humidity control (Fig. 3.7 to 3.9). As shown in Fig. 3.19, the internal relative humidity of concrete increases as the temperature increases. However, the increase of the internal relative humidity is very small (3 % at temperature 28 C and 8 %, the peak value at temperature 77 C). If the increase of the internal relative humidity during heating is neglected the strain measured from the dried specimen can be regarded as PTS. The PTS is potted versus temperature in Fig. 3.21. As shown in Fig. 3.21, the PTS is linear because there is no moisture effect, unlike the CTS. For temperature ranging from 28 C to 70 C, the function of the PTS is expressed with Eq. (3.15). T is the temperature in Celsius.

109

PTS = (10.995 T 307.86 ) 106

for 28 C T 70 C

(3.15)

100 90 80 RH (%) & Temp (C) 70 60 50 40 30 20 10 0 0 50 100 150 Time (hrs) 200 250 300
Temp (Chamber) RH (Chamber)

Figure 3.18. Temperature and RH inside the chamber vs. Time

100 90 80 RH (%) & Temp (C) 70 60 50 40 30 20 10 0 0 50 100 150 Time (hrs) 200 250 300
Temp (Specimen) RH (Specimen) Strain_average

1200 1000 800 Strain (x10-6) 600 400 200 0 -200 -400

Figure 3.19. Temperature, RH and strain history of the specimen vs. Time

110

100 90 80 RH (%) & Temp (C) 70 60 50 40 30 20 10 0 0 50 100 150 Time (hrs) 200

Temp (Chamber) RH (Chamber) Temp (Specimen) RH (Specimen) Strain_average

1200 1000 800 Strain (x10-6) 600 400 200 0 -200 -400

250

300

Figure 3.20. RH, temperature, and strain history vs. time (Dried specimen)

600
PTS

500

PTS Trend

400 Strain (x10-6)

300

200

100

0 20 30 40 50 Temperature (C) 60 70 80

Figure 3.21. Pure thermal strain by temperature increase from 28 C to 70 C

111
3.3.3. Thermo-hygro coupling effect

From experimental data, the thermo-hygro coupling effect can be obtained using Eq. (3.16).

CTS = PTS + PHS + thermo - hygro coupling ( coupling )

(3.16)

To plot the thermo-hygro coupling effect ( coupling ) on the plane of strain versus temperature with CTS and PTS , the factor f is defined as Eq. (3.17).

f =
Thus, from Eq. (3.12),

( Pa Pw ) vw
R

(3.17)

T In ( RH ) = f

(3.18)

The factor f of Range A, where both temperature and internal humidity of concrete vary, in Fig. 3.9 is calculated using Eq. (3.18). To find an equivalent RH at a constant temperature of 77.5 C corresponding to Range A in Fig. 3.9, the calculated

f and temperature 77.5 C are placed into Eq. (3.18). In the equation, T is the
temperature in Kelvin. Fig. 3.22 shows the internal relative humidity and temperature in Kelvin versus the factor f for concrete in Range A of Fig. 3.9. Fig. 3.23 is the equivalent

112 relative humidity at constant temperature 77.5 C (350.65 Kelvin) versus the factor

f , which is the same as that in Fig 3.22.

100 90 80
RH (%) Temperature (Kelvin)

360 350 Temperature (Kelvin) 340 330 320 310 300 290 1600

70 RH (%) 60 50 40 30 20 10 0 1200

1250

1300

1350

1400 f (factor)

1450

1500

1550

Figure 3.22. RH (%) and Temp. (Kelvin) versus factor (f) for the range A in Fig. 3.9

100 90 80 70 RH (%) 60 50 40 30 20 10 0 1200 1250 1300 1350

Equivalent RH (%) at 350.65 Kelvin Temperature (350.65 Kelvin)

360 350 Temperature (Kelvin) 340 330 320 310 300 290 1600

1400 f (factor)

1450

1500

1550

Figure 3.23. Equivalent RH (%) at Temp. 77.5C (350.65 Kelvin) versus factor (f)

113 Now, thermo-hygro coupling effect ( coupling ) can be plotted on the plane of strain versus temperature with CTS and PTS because the PHS at constant temperature 77.5 C (350.65 Kelvin) was already obtained in the relation of strain versus RH (%) as shown in Fig. 3.15. The thermo-hygro coupling strain with CTS, PTS, and PHS in the temperature range between 28-70 C is shown in Fig. 3.24.

600 500 400 Strain (x10-6) 300 200 100 0 -100 20 30 40 50 Temperature (C) 60 70 80
CTS PTS PHS Thermo-hygro coupling strain

Figure 3.24. Thermo-hygro coupling strain between 28-70C

In Fig 3.24, PTS is less than CTS because the there is no additional hygro strain as the temperature increases. PHS is nonlinear. At higher temperatures, PHS increases more rapidly. The thermo-hygro coupling strain is negative in the tested temperature range for this study. This means that there is no shrinkage due to

114 dehydration (decrease of relative humidity) in the CTS of concrete. The thermo-hygro coupling strain from curve fitting is expressed with Eq. (3.19). T is the temperature in Celsius.

coupling = 23.275 1010 T 3 + 30.541 108 T 2

13.151 106 T + 179.877 106

for 28 C T 70 C

(3.19)

3.4. Conclusions

The Conventional Thermal Strain (CTS), Pure Thermal Strain (PTS), and Pure Hygro Strain (PHS) with respect to temperature and humidity changes were measured continually and simultaneously over time. From the measured strains, the thermohygro coupling effect in the temperature range 28-70 C was obtained. The conclusions from this study are as follows; 1. If it is assumed that a local thermodynamic equilibrium always exists between the phases of pore water (vapor, liquid) within a very small element of concrete, the phenomenon, which is the increase of internal relative humidity of concrete with temperature increase, can be explained using Eq. (3.12). This equation is based on static force equilibrium from a capillary tube model and physicochemical equilibrium by the Kelvin equation. When we consider the strain change of aggregates and hardened cement paste as components of concrete at elevated temperatures, it is clear that the internal relative humidity of concrete increases as the temperature increases.

115 2. Pure Thermal Strain (PTS), because the there is no additional moisture effect by swelling pressure increase (increase of relative humidity) according to temperature increase, was less than Conventional Thermal Strain (CTS) in the range of temperatures used in this study. 3. Pure Hygro-Strain (PHS) was nonlinear. At higher temperatures, Pure Hygro Strain (PHS) increased more rapidly. It is likely linked to complicated physiochemical reactions in concrete as humidity changes, i.e. the increase of swelling pressure (RH increase) is not directly proportional to temperature increase. 4. The thermo-hygro coupling effect was negative in the tested temperature range. That means that the internal relative humidity of concrete increases as the temperature increases in the tested temperature range. 5. The shrinkage effect due to a decrease in internal relative humidity (dehydration) of cement may be contained in the overall expansion of concrete beyond 150 C. However, it should be noticed that concrete expands as temperature increases because aggregates, which make up about 70 % of concrete by volume, expand continually as the temperature increases.

116

CHAPTER 4

4.

STRENGTH AND DURABILITY OF CONCRETE SUBJECTED TO VARIOUS HEATING AND COOLING TREATMENTS

4.1. Introduction

Under rapid heating, concrete experiences large volume changes resulting from thermal dilatation of aggregate, shrinkage of cement paste due to moisture loss, and spalling damage due to high thermal stresses and pore pressure build-up. Therefore, it is very important to study residual mechanical and durability properties after concrete is subjected to high temperatures with different heating and cooling rates. These special features of concrete will provide essential information to the concrete industry for improving the fire resistance of concrete. Extensive experimental studies on this important topic were performed in the past. The important experimental parameters included maximum temperature, heating rate, types of aggregates used, various binding materials, and mechanical loads under high temperature conditions. However, the experimental studies have mainly concentrated on the strength of concrete, very little research focused on the reduction of concrete durability caused by fire damage. Poon et al. (2001) conducted an

117 experimental study on the strength and durability performance of normal and high strength concretes, but the study did not consider the effects of heating and cooling rates which are very important factors on degradation of concrete due to fire. From a strength point of view, the reduction of concrete strength is affected strongly by the heating rate, and the residual strength of concrete depends strongly on the cooling rate. The purpose of this experimental study is to investigate both the strength and durability performance of concrete subjected to various high temperature scenarios. The test variables are heating rates, holding times at target temperatures, and cooling rates. Thermal diffusivity, which is one of the fundamental thermal properties of concrete, is evaluated after various heat treatments. The reduction of the durability of concrete is investigated using a water permeability test (WPT). The unstressed residual compressive strength test and ultrasonic pulse velocity test (UPV) are performed to investigate the strength and stiffness deterioration by each high temperature scenario. Additionally, weight losses, color changes, and cracks of the specimens are also studied and reported.

4.2. Specimen preparation, heating equipment and test variables

The mix proportion and materials used to produce the concrete specimens are the same as that previously described in chapter 3.2.1. Concrete cylindrical specimens with dimensions of 48 inch are made for the experimental study. The maximum size of coarse aggregate in concrete specimen is 3/4 inch. The specimens are cured under the standard condition (Temperature 23 C and RH 93 %) for 8 weeks in a fog room.

118 For the weight loss and water permeability tests 42 inch slices are needed. To make the slices the concrete cylinders are cured for 21 days in the fog room and then cut into the slices. To make the cuts from 48 inch cylinders the bottom and top 1.5 inches of the specimens are removed, the remaining portion of the cylinders is cut in half, and then put into the fog room for another five weeks of curing. An electrically heated furnace (Model RHF 15/8 of CARBOLITE) designed for maximum temperature up to 1600 C is used. The interior size of the furnace is 7.8"W11.4"D6.6"H. The temperature history inside the furnace is measured and recorded from type K-thermocouples installed inside. The data logger used along with the thermocouples is the model OM-CP-OCTTEMP produced by OMEGA. The water permeability test is performed after the specimen experiences one of the temperature scenarios. The test variables are heating rates (2 and 15 C/min), holding times (2 and 4 hrs) at each maximum temperature and cooling conditions, which include slow cooling (1 C/min), natural cooling in the furnace and water cooling. The target maximum temperatures are 200, 400, and 600 C. For the unstressed residual compressive strength test and ultrasonic pulse velocity test (UPV), the test variables are the cooling condition (slow cooling, natural cooling, and water cooling) and maximum temperatures 200, 400, 600, and 800 C. The heating rate is 2 C/min, and holding time at the maximum temperature is 4 hrs. In the cases of natural and slow cooling, the specimen is left in the furnace and the temperature change is recorded over time. For water cooling, the specimen is taken out of the furnace and put in a tank of water that is initially at 20 C. Fig. 4.1 is the schematic of the temperature scenarios used for this test.

119

Figure 4.1. Schematic of temperature scenarios

Mullite

Thermocoup

Figure 4.2. Test set-up for heating

The furnace is heated using two groups of three heating elements on either side of the specimen. In order to protect the heating units from potential spalling damage of concrete, the concrete specimen is placed inside a hollow tube made of mullite with a maximum operational temperature of 1700 C. This provided a more uniform thermal condition inside the specimen. Fig. 4.2 shows the test-set up for heating. Fig 4.3 shows the temperature histories measured in the mullite tube with the K-thermocouple.

120

700
R15_200D2S1

600 500 Temperature (C) 400 300 200 100 0 0 200 400 600 800 Time (min) 1000 1200

R15_200D2N R15_200D2W

Water cooling

Slow cooling Natural cooling

1400

1600

(a) Heating rate 15 C/min, target temperature 200 C, and different cooling regimes

700
R15_400D2S1

600 500 Temperature (C) 400 300 200

Natural cooling Slow cooling Water cooling

R15_400D2N R15_400D2W

100 0 0 200 400 600 800 Time (min) 1000 1200 1400 1600

(b) Heating rate 15 C/min, target temperature 400 C, and different cooling regimes

121

700 600 500 Temperature (C) 400 300 200 100 0 0

Water cooling

R15_600D2S1 R15_600D2N R15_600D2W

Slow cooling Natural cooling

200

400

600

800 Time (min)

1000

1200

1400

1600

(c) Heating rate 15 C/min, target temperature 600 C, and different cooling regimes

700
R2_200D4S1

600 500 Temperature (C) 400 300 200 100 0 0 200 400 600 800 Time (min) 1000 1200

R2_200D4N R2_200D4W

Water cooling

Slow cooling Natural cooling

1400

1600

(d) Heating rate 2 C/min, target temperature 200 C, and different cooling regimes

122

700
R2_400D4S1

600 500 Temperature (C) 400 300 200 100 0 0 200 400 600 800
Time (min)

R2_400D4N R2_400D4W Water cooling

Slow cooling Natural cooling

1000

1200

1400

1600

(e) Heating rate 2 C/min, target temperature 400 C, and different cooling regimes

700
Water cooling R2_600D4S1 R2_600D4N R2_600D4W

600 500 Temperature (C) 400 300 200 100 0 0 200 400 600 800 Time (min) 1000 1200
Slow cooling

Natural cooling

1400

1600

(f) Heating rate 2 C/min, target temperature 600 C, and different cooling regimes Figure 4.3. Temperature histories measured at the inside of the mullite tube

123
4.3. Temperature distribution and thermal diffusivity

4.3.1. Test set-up

Thermal diffusivity is one of the fundamental thermal properties, such as gas or liquid permeability, used to evaluate the bearing capacity of a fire-exposed concrete structure. The thermal diffusion is evaluated with a 48 concrete cylinder.

(a) Specimen geometry

(b) Positions of thermocouples

(c) Set-up of thermal couples (Type K)

(d) Specimen placed inside the furnace

Figure 4.4. Specimen geometry, locations of thermocouples and test set-up

124

One thermocouple is used to measure ambient temperature between the specimen and mullite hollow tube, and four additional K thermocouples are installed inside and on the surface of the specimen to measure the temperature distribution in the concrete cylinder. Fig. 4.4 shows the specimen geometry, locations of thermocouples and test set-up. The heating rate is 1 C/min, maximum temperature 900 C, holding time at the maximum temperature 2 hrs, and the cooling method is natural cooling.

4.3.2. Results and discussion

1000 900 800 700 Temperature (C) 600 500 400 300 200 100 0 0 5 10 15 20 Time (hrs) 25 30 35 40
Position 1 Position 2 Position 3 Position 4 Ambient

Figure 4.5. Transient temperature at each position

125 Fig. 4.5 shows temperature histories over time measured at each thermocouple location. One can see that in the heating phase the surface temperature is higher than the internal temperature and in the cooling phase the surface temperature is lower than the internal temperatures, which are expected experimental results. The two vertical drops in the cooling phase in Fig. 4.5, near 540 C at the center and 330 C at mid-radius, are due to the damage of the two thermocouples in position 1 and position 2.

70 60 50 Delta T (C) 40 30 20 10 0 0 200 400 600 800 1000 Surface temperature (C)
The first peak The third peak

The second peak

Figure 4.6. T Vs. Surface temperature

An interesting result in Fig. 4.5 is the radial distribution of temperatures measured at different locations during the testing period. Fig. 4.6 shows the temperature difference between the surface and center of the specimen at different temperature ranges. One can see that the temperature difference varies during the

126 entire testing period, which means that a steady state condition in the concrete was never reached. This is mainly due to the phase transformations taking place at different temperatures in the concrete. The three peaks shown in the figure are related to the micro-structural changes due to complex physicochemical transformations in the concrete under different high temperatures. The peaks are closely related to the results of DTA (Differential Thermal Analysis) conducted by Lankard (1970). The first peak occurs at 200 C. This peak is associated with the evaporation of free water and the dehydration of calcium silicate hydrate (C-S-H). In fact, the weight of the concrete is decreased rapidly up to about 200 C due to the evaporation of free water. This result is shown more clearly in the weight loss measured in this study, which will be discussed later. The second peak between 550 C and 650 C is related to the decomposition of calcium hydroxide (CH) and calcium silicate hydrate (C-S-H). The peak occurring at 850 C is occasionally observed in DTA. It is likely that the third peak is due to the decomposition of calcium carbonate ( CaCO3 ) observed in the DTA test results by Lankard (1970). Fig. 4.7 shows the temperature profiles measured at positions 1, 2, and 3, when temperatures at the center in the specimen have reached about 600 C and 800 C, during both heating and cooling. The surface temperature corresponding to each temperature profile is indicated in the figure. While undergoing natural cooling, the temperature difference between the center and the surface is two times larger than that during heating. This means that the damage of the concrete is strongly affected by cooling method. One of the important applications of the temperature profiles is to determine the thermal diffusivity of the cylindrical concrete specimens. The

127 simplified expression, developed by Khoury et al (1985 a), for linear heat conduction in a solid cylinder subjected to constant surface temperature increase relates the thermal diffusivity ( D ) to the temperature difference ( T ) between the surface and the center. It does not account for moisture transport.

D = vh R 2 / 4T

(4.1)

vh is the rate of temperature of the specimen at the surface and R is the radius of the specimen.

680

630 600.22

During heating [Surface Temp: 1179.66 F (637.59 C)] 636.86 636.86

575.49

575.49 550.55

609.43 600

609.43

550.55 During cooling [Surface Temp: 976.98 F (524.99 C)] 530 -40.8

580 -40.8

-20.4

20.4

40.8

-20.4

20.4

40.8

890
During heating [Surface Temp: 1572.75 F (855.97 C)] 845.49 822.75 811.06 822.75 845.49

830 812.21 796.15 796.15

During cooling [Surface Temp: 1302.01 F (705.56 C)] 741.48 741.48 -20.4 0 20.4 40.8

790 -40.8

-20.4

20.4

730 40.8 -40.8

Figure 4.7. Temperature profiles during heating and cooling

Fig. 4.8 is the plot of thermal diffusivity versus temperature based on the temperature profiles. One can see that there is a large drop in thermal diffusivity up to about 200 C, which is related to the loss of the free water, and it remains relatively constant thereafter with increasing temperature.

128

2.5

D*1000 (m2/h)

1.5

0.5

0 0 100 200 300 400 500 600 700 800 900

Temperature (C)

Figure 4.8. Thermal diffusivity

4.4. Water permeability

4.4.1. Test set-up

Water permeability is a measurement of the transport property of concrete, and a very good indicator of the change of microstructure in concrete under different temperature treatments. As shown in this section, the permeability is far more sensitive to temperature treatment than the associated degradation of strength and stiffness which will be discussed in the subsequent section. Figs. 4.9 and 4.10 show the procedure of the test sep-up and the experimental apparatus used in the present study for measuring water permeability of concrete,

129 respectively [see Ludirdja et al. (1989)]. The flow of water is measured from the drop of water level in the pipette during the test. The water level is filled to the original level periodically. After the specimen is exposed to a temperature cycle, it is placed in an air vacuum for 3 hours to remove air in the specimen, and then it is placed in a water vacuum for 1 hour. The water used for the water vacuum is boiled to make sure it is free of air and then it is cooled to room temperature. The specimens are then immersed for 18 hrs in the de-aerated water and finally placed in the water permeability apparatus. The vacuum process follows ASTM C 1202.

(a) Put specimen into plastic tube

(b) Combine top and bottom cover

(c) Put water into bottom and top cover

(d) Test set-up

Figure 4.9. Procedure of the test sep-up

130

Figure 4.10. Test set-up detail

4.4.2. Results and discussion

The test results for the test series R15_D2 (heating rate15 C/min and 2 hrs holding time at target temperature) are plotted in Fig. 4.11 as the cumulative water permeated versus time. In Fig. 4.11, the plot (a), (b) and (c) show the test results subjected to different target temperatures under the same cooling regime. The plots (d), (e), and (f) show the effect of different cooling regimes for the same target temperature. The test results for the test series R2_D4 (heating rate 2 C/min and 4 hrs holding time at target temperature) are plotted in Fig. 4.12 with same format as those of the test series R15_D2.

131

60
Reference (No heating) R15_200D2S1 R15_400D2S1 R15_600D2S1

60
Reference (No heating) R15_200D2N

50

50

R15_400D2N R15_600D2N

40 Q (ml)

40 Q (ml)
0 2000 4000 6000 8000 10000 12000 14000

30

30

20

20

10

10

0 Time (min)

0 0 2000 4000 6000 8000 10000 12000 14000 Time (min)

(a) Slow cooling

60
Reference (No heating) R15_200D2W

(b) Natural cooling

60
Reference (No heating) R15_200D2S1 R15_200D2N R15_200D2W

50

R15_400D2W R15_600D2W

50

40 Q (ml)
Q (ml)

40

30

30

20

20

10

10

0 0 2000 4000 6000 8000 10000 12000 14000 Time (min)

0 0 2000 4000 6000 8000 10000 12000 14000 Time (min)

(c) Water cooling

60
Reference (No heating) R15_400D2S1

(d) Maximum temperature 200 C

60
Reference (No heating) R15_600D2S1

50

R15_400D2N R15_400D2W

50

R15_600D2N R15_600D2W

40 Q (ml) Q (ml)

40

30

30

20

20

10

10

0 0 2000 4000 6000 8000 10000 12000 14000 Time (min)

0 0 2000 4000 6000 8000 10000 12000 14000 Time (min)

(e) Maximum temperature 400 C

(f) Maximum temperature 600 C

Figure 4.11. Cumulative water permeated versus time for the test series R15_D2

132

60
Reference (No heating) R2_200D4S1 R2_400D4S1 R2_600D4S1

60
Reference (No heating) R2_200D4N R2_400D4N R2_600D4N

50

50

40 Q (ml) Q (ml) 0 2000 4000 6000 8000 10000 12000 14000

40

30

30

20

20

10

10

0 Time (min)

0 0 2000 4000 6000 8000 10000 12000 14000 Time (min)

(a) Slow cooling

60
Reference (No heating) R2_200D4W R2_400D4W R2_600D4W

(b) Natural cooling

60
Reference (No heating) R2_200D4S1 R2_200D4N R2_200D4W

50

50

40 Q (ml) Q (ml) 0 2000 4000 6000 8000 10000 12000 14000

40

30

30

20

20

10

10

0 Time (min)

0 0 2000 4000 6000 8000 10000 12000 14000 Time (min)

(c) Water cooling

60
Reference (No heating) R2_400D4S1 R2_400D4N R2_400D4W

(d) Maximum temperature 200 C

60
Reference (No heating) R2_600D4S1 R2_600D4N R2_600D4W

50

50

40 Q (ml) Q (ml) 0 2000 4000 6000 8000 10000 12000 14000

40

30

30

20

20

10

10

0 Time (min)

0 0 2000 4000 6000 8000 10000 12000 14000 Time (min)

(e) Maximum temperature 400 C

(f) Maximum temperature 600 C

Figure 4.12. Cumulative water permeated versus time for the test series R2_D4

133 The nonlinearity of the curves may be attributed to the absorption of the water into a partially empty pore system in the specimens. The test results show that the initial slope of cumulative water flow through the specimen increases as the target temperature and cooling rate increase. There is linearity in the plots of the specimens exposed to a maximum temperature over 400 C using water cooling. This may be due to the severe damage caused by rapid cooling. These trends are also shown in the test series R2_D4.

The coefficient of permeability is obtained from Darcys law: V = K p Ah / L By integrating the above equation, Q L /( A h) = K p t (4.3) (4.2)

Where V is the rate of flow dQ dt per unit area ( m / sec ), Q is the total water permeated ( m 3 ), A is the cross section of the specimen ( m 2 ); L is the thickness ( m ), h is the head of water ( m ), and K p is the coefficient of permeability ( m / sec ). Note that A and L in our specimens are constants, and h is also assumed to be constant if the small difference in pressure head during the test is neglected.

Fig. 4.13 (Test series R15_D2) and 4.14 (Test series R2_D4) are the plots of linear regression of Q L /( A h) versus time after the steady state flow is nearly established. The coefficient of permeability K p is calculated from the slope found from the linear regression.

134
0.0003 Regression functions REF (No heating) R15_200D2S1 0.00025 R15_200D2N R15_200D2W R15_400D2S1 0.0002 R15_400D2N R15_400D2W R15_600D2S1 y = 7.70E-09x - 4.11E-07 y = 7.98E-09x - 6.14E-07 y = 1.03E-08x + 1.70E-06 y = 1.49E-08x + 4.17E-06 y = 1.82E-08x - 1.41E-07 y = 1.89E-08x + 1.40E-06 y = 1.42E-06x + 2.36E-05 y = 2.68E-08x + 1.09E-05 y = 3.37E-08x + 1.59E-05 y = 1.45E-05x + 4.13E-07

QL/Ah (m)

R15_600D2N 0.00015 R15_600D2W

0.0001

0.00005

0 0 1000 2000 3000 4000 5000 6000 7000

Time (min)

Figure 4.13. Linear regression for the test series R15_D2

0.0003 Regression functions Ref (No heating) 0.00025 R2_200D4S1 R2_200D4N R2_200D4W R2_400D4S1 0.0002 R2_400D4N R2_400D4W y = 7.70E-09x - 4.11E-07 y = 8.64E-09x - 4.36E-07 y = 1.09E-08x + 3.84E-06 y = 2.24E-08x + 1.16E-05 y = 1.39E-08x + 3.66E-06 y = 1.53E-08x + 3.75E-06 y = 2.63E-06x + 1.66E-05 y = 1.89E-08x + 1.63E-05 y = 2.31E-08x + 1.58E-05 y = 1.97E-05x - 4.03E-19

QL/Ah (m)

R2_600D4S1 0.00015 R2_600D4N R2_600D4W

0.0001

0.00005

0 0 1000 2000 3000 4000 5000 6000 7000 8000 9000

Time (min)

Figure 4.14. Linear regression for the test series R2_D4

135

Table 4.1 and 4.2 show the values of the coefficient of permeability for the test series R15_D2 and R2_D4 respectably.

Table 4.1 Coefficients of water permeability for R15_D2 test series Temp (C) 25 C 200 C 400 C 600 C Slow cooling Kp, (10-10 m/s) 1.28 1.33 3.03 4.47 Natural cooling Kp, (10-10 m/s) 1.28 1.72 3.15 5.62 Water cooling Kp, (10-10 m/s) 1.28 2.48 236.67 2416.67

Table 4.2 Coefficients of water permeability for R2_D4 test series Temp (C) 25 C 200 C 400 C 600 C Slow cooling Kp, (10-10 m/s) 1.28 1.44 2.32 3.15 Natural cooling Kp, (10-10 m/s) 1.28 1.82 2.55 3.85 Water cooling Kp, (10-10 m/s) 1.28 3.73 438.33 3283.33

Effects of the maximum temperature and cooling rate

The results shown in Table 4.1 and Table 4.2 indicate that the value of K p increases when the specimen is subjected to higher temperatures and faster cooling rates. In terms of the development of damage in concrete, it increases as the maximum temperature and cooling rate increase, which leads to an increase in K p . Increasing the cooling rate affects the level of damage to the concrete because of the

136 increasing temperature gradients. Particularly, the K p values of the specimens subjected to the target temperature 400 or 600 C under water cooling are very high when compared with those subjected to other cooling regimes (natural cooling and slow cooling) as shown in Table 4.1 and Table 4.2. Column 1 of Table 4.1 and Table 4.2 show that under slow cooling, the permeability increases very little. The last column of the two tables shows the effect of fast cooling where a very large increase in permeability can be observed. This is evidence that permeability is very sensitive to the variation of the microstructure of concrete, which changes depending on the applied temperature treatment. The cooling rate appears to be a significant factor in the damage development of concrete which was exposed to high temperatures.

5.0
R15D2_Slow cooling R15D2_Natural cooling 4.38

4.0 Relative water permeability

R15D2_Water cooling 3.48

3.0
2.45 2.36

2.0

1.94 1.34

1.0

1.00

1.04

0.0 25 200 Temperature (C) 400 600

Figure 4.15. Relative water permeability (test series R15_D2)

137

5.0
R2D4_Slow cooling R2D4_Natural cooling R2D4_Water cooling

4.0 Relative water permeability

3.0

2.91

3.00 2.45

2.0
1.42 1.12

1.99 1.81

1.0

1.00

0.0 25 200 Temperature (C) 400 600

Figure 4.16. Relative water permeability (test series R2_D4)

Fig. 4.15 and 4.16 are the plots of relative water permeability versus maximum temperature for the test series R15_D2 and R2_D4, respectively. These figures clearly show the effects of target temperature and cooling rate.

Effects of the holding time at target temperature and heating rate

The comparisons between test series R2_D4 and R15_D2 are shown in Fig. 4.17 and Fig. 1.18. For the specimens exposed to a target temperature over 400 C (Fig. 4.17 and Fig. 1.18), the effect of the heating rate is more significant than the holding time. Under lower temperature ranges, up to a target temperature of 200 C, the difference between the two effects is not as significant.

138

5.0 4.5 4.0 Relative water permeability 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0 25 200 Temperature (C) 400 600 1.00 1.12 1.04 2.36 1.81 2.45 3.48
R2D4_Slow cooling R15D2_Slow cooling

Figure 4.17. Comparison between R2_D4 and R15_D2 subjected to slow cooling

5.0 4.5 4.0 Relative water permeability 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0 25 200 Temperature (C) 400 600 1.00 1.42 1.34 2.45 1.99 3.00
R2D4_Natural cooling R15D2_Natural cooling

4.38

Figure 4.18. Comparison between R2_D4 and R15_D2 subjected to natural cooling

139
4.5. Ultrasonic Pulse Velocity (UPV) and residual compression test

4.5.1. Test equipments and test set-up

For this test, 48 inch cylinders are used. An ultrasonic pulse velocity (UPV) test is performed prior to the residual compression test. The equipment for UPV is VMETER from JAMES INSTRUMENTS. The longitudinal wave pulse velocity determination method is popular for nondestructive testing of concrete because of its simplicity and cost effectiveness.

Figure 4.19. Test set-up of the residual compression test

There are three standards in arranging transducers, which are direct transmission (located direct opposite to each other), diagonal transmission (located

140 diagonally to each other), and indirect transmission (located on same surface and separated by known distance). The direct transmission method is used in this research because the direct method is the most sensitive among the three methods (Komlo at al, 1996). In residual compression test, axial deformation is measured with MTS model 632.94E-20, which delivers the average value automatically from two extensometers. Displacement control of 0.0001 inch/sec is used for the test. Fig. 4.19 shows the setup of the residual compression test. As mentioned earlier, the test variables are the maximum temperature exposed and cooling regimes with constant heating rate of 2 C/min and holding time of 4 hrs at the maximum temperatures. The target temperatures are 200, 400, 600, and 800 C.

4.5.2. Results and discussion for UPV (Ultrasonic pulse velocity)

Table 4.3 shows the results from ultrasonic pulse velocity test (UPV). The pulse frequency used for testing concrete is much lower than that used in metal testing. The 54 kHz transducers are used for concrete testing. The signal wave length is about 3 inches. Fig. 4.20 is a plot for relative velocity versus maximum temperature exposed. As shown in the test results, the pulse velocity decreases as the exposed maximum temperature increase for all cases of cooling methods. In other words, the level of damage is fairly proportional to maximum temperature. In the subject of cooling rate, it is observed that the velocity of specimens subjected to slow cooling are higher than that of specimens subjected to natural cooling. This implies that faster

141 cooling contributes to increase of damage. However, this tendency does not apply to the comparison of the water cooling to slow or natural cooling because the specimens were temporarily (about 10 min) submerged in the water, and some water penetrates to pores and cracks of the specimen. Thus, the test results from the specimens subjected to water cooling can not be an accurate indication of level of damage.

Table 4.3. Ultrasonic pulse velocity test result Temp. (C) 25 200 400 600 800 Velocity (m/sec) Natural cooling 3.97E+03 3.20E+03 2.36E+03 1.01E+03 3.81E+02

Slow cooling 3.97E+03 3.23E+03 2.43E+03 1.10E+03 6.55E+02

Water cooling 3.97E+03 3.03E+03 2.22E+03 1.17E+03 6.77E+02

1.2
R2D4_Slow cooling

1.0

R2D4_Natural cooling R2D4_Water cooling

Relative velocity

0.8

0.6 0.4

0.2

0.0 25 200 400 Temperature (C) 600 800

Figure 4.20. Relative velocity versus maximum temperature exposed

The elastic modulus is calculated using the test result of pulse velocity. The function is shown below:

142
2 Vlong (1 + ) (1 2 )

Ep =

(1 )

(4.4)

Where E p is elastic modulus, Vlong is longitudinal pulse velocity (P-wave), is density of the solid material, and is Poissons ratio of the solid material (0.2).

Table 4.4. Unit weight measured from specimens used in present study Temp. (C) 25 200 400 600 800 Unit weight ( kN m 3 ) Slow cooling Natural cooling Water cooling 2.304 2.304 2.184 2.178 2.304 2.162 2.158 2.129 2.133 2.098 2.098

Table 4.5. Elastic modulus ( E p ) from ultrasonic pulse velocity Temp. (C) 25 200 400 600 800 E p (psi) Slow cooling 4.75E+06 2.97E+06 1.67E+06 3.37E+05 1.18E+05 Natural cooling 4.75E+06 2.91E+06 1.57E+06 2.81E+05 3.96E+04 Water cooling 4.75E+06 2.76E+06 1.49E+06 4.10E+05 1.38E+05

Table 4.4 is a summary for unit weights measured from concrete samples used in present study. In the calculation of elastic modulus, the unit weights in Table 4.4 are used. In case of the water cooling, the unit weight measured from no heating specimens is used. Table 4.5 shows the elastic modulus calculated from the ultrasonic pulse velocity. Fig. 4.21 is a chart of relative elastic modulus (obtained from UPV) versus maximum temperature. The elastic modulus obtained from ultrasonic pulse

143 velocity (UPC) is compared to initial tangent modulus obtained from the residual compression test in section 4.5.4.

1.2 R2D4_Slow cooling 1.0 Relative dynamic modulus 0.8 0.6 0.4 0.2 0.0 25 200 400 Temperature (C) 600 800 R2D4_Natural cooling R2D4_Water cooling

Figure 4.21. Relative elastic modulus (from UPV) versus maximum temperature

4.5.3. Results and discussion for residual compression test

Fig 4.22, 4.23, and 4.24 plot stress versus strain for the specimens that were subjected to slow, natural, and water cooling, respectively. The compressive strength and elastic modulus decrease significantly as the maximum temperature increases. The elastic region of the specimens subjected to slow or natural cooling and temperatures of 600 C and beyond are nonlinear. For the case of water cooling, the same trend is shown for the range of maximum temperatures of 400 C and beyond. This may be due to thermally induced cracks formed during high temperature treatment. In the initial stage of loading the cracks are closed as the load gradually

144 increases. After the cracks are closed, the curves follow the general trend of a compression strength test.

3500 3000 2500 Stress (psi) 2000 1500 1000 500 0 0 0.005 0.01 0.015 Strain (in/in) 0.02 0.025 0.03
Ref (No heating) R2_200D4S1 R2_400D4S1 R2_600D4S1 R2_800D4S1

Figure 4.22. Stress versus strain (Slow cooling)

3500 3000 2500 Stress (psi) 2000 1500 1000 500 0 0 0.005 0.01 0.015 Strain (in/in) 0.02 0.025 0.03
Ref (No heating) R2_200D4N R2_400D4N R2_600D4N R2_800D4N

Figure 4.23. Stress versus strain (Natural cooling)

145

3500 3000 2500 Stress (psi) 2000 1500 1000 500 0 0 0.005 0.01 0.015 Strain (in/in) 0.02 0.025 0.03
Ref (No heating) R2_200D4W R2_400D4W R2_600D4W R2_800D4W

Figure 4.24. Stress versus strain (Water cooling)

3500
Ref (No heating)

3000 2500 Stress (psi) 2000 1500 1000 500 0 0 0.005 0.01 0.015 Strain (in/in) 0.02

R2_200D4S1 R2_200D4N R2_200D4W

0.025

0.03

Figure 4.25. Stress versus strain (Maximum temperature 200C)

146

3500
Ref (No heating)

3000 2500 Stress (psi) 2000 1500 1000 500 0 0 0.005 0.01 0.015 Strain (in/in) 0.02

R2_400D4S1 R2_400D4N R2_400D4W

0.025

0.03

Figure 4.26. Stress versus strain (Maximum temperature 400C)

3500
Ref (No heating)

3000 2500 Stress (psi) 2000 1500 1000 500 0 0 0.005 0.01 0.015 Strain (in/in) 0.02

R2_600D4S1 R2_600D4N1 R2_600D4W

0.025

0.03

Figure 4.27. Stress versus strain (Maximum temperature 600C)

147

3500
Ref (No heating)

3000 2500 Stress (psi) 2000 1500 1000 500 0 0 0.005 0.01 0.015 Strain (in/in) 0.02

R2_800D4S1 R2_800D4N R2_800D4W

0.025

0.03

Figure 4.28. Stress versus strain (Maximum temperature 800C)

The effect of different cooling methods is shown in Fig. 4.25 through Fig. 4.28. Faster cooling contributes to decrease in strength and elastic modulus of the concrete as shown in Figs.

Table 4.6. Ultimate strength results of residual compression test Temp. (C) 25 200 400 600 800 f c' : ultimate strength (psi) Slow cooling 3068.59 3067.76 2856.25 1734.85 538.65 Natural cooling 3068.59 3011.87 2695.97 1554.84 463.03 Water cooling 3068.59 2221.44 1978.69 1163.05 425.22

Table 4.6 is a summary of the ultimate strengths from the residual compression test. Table 4.7 and Fig. 4.29 show the residual strength normalized by

148 the reference strength (25 C) for three cooling methods as a function of the target temperature. The residual strength rapidly drops beyond 400 C. The strength of the specimens subjected to water cooling decreases more rapidly than the specimens subjected to other cooling methods. The strength of the specimens exposed to the target temperature of 600 C is less than 57 % of the reference strength. For 800 C, the strength is less than 18 % of the reference strength.

Table 4.7. Relative ultimate strength Temp. (C) 25 200 400 600 800 Relative ultimate strength Slow cooling Natural cooling Water cooling 1 1 1 0.9997 0.9815 0.7239 0.9308 0.8786 0.6448 0.5654 0.5067 0.3790 0.1755 0.1509 0.1386

1.2
R2D4_Slow cooling

1.0 0.8 0.6 0.4 0.2 0.0 25 200 400 Temperature (C) 600

R2D4_Natural cooling R2D4_ Water cooling

Relative strength

800

Figure 4.29. Relative residual strength vs. maximum temperature

149 Table 4.8 summarizes the test results for the initial tangent modulus from the residual compression test.

Table 4.8. Initial tangent modulus from residual compression test Temp.(C) 25 200 400 600 800 Ei : initial tangent modulus (psi) Slow cooling 3.60E+06 2.34E+06 1.14E+06 1.91E+05 4.39E+04 Natural cooling 3.60E+06 2.09E+06 1.04E+06 1.38E+05 2.44E+04 Water cooling 3.60E+06 1.63E+06 3.53E+05 5.77E+04 1.30E+04

Table 4.9 and Fig. 4.30 show the values of the initial tangent modulus with different temperatures, normalized by the reference specimen tested at room temperature. The basic trend of the variation of initial tangent modulus shown in Fig. 4.30 is similar to that of the residual strength shown in Fig. 4.29. However, the initial tangent modulus of concrete is more sensitive to elevated temperature than the compression strength. The initial tangent moduli of the specimens exposed to a maximum temperature of 600 C are less than 5.3 % of the initial reference tangent modulus. For 800 C, the moduli are less than 1.3 % of the reference values.

Table 4.9. Relative initial tangent modulus Temp. (C) 25 200 400 600 800 Relative initial tangent modulus Slow cooling Natural cooling Water cooling 1.0000 1.0000 1.0000 0.6510 0.5816 0.4537 0.3183 0.2896 0.0983 0.0530 0.0385 0.0161 0.0122 0.0068 0.0036

150
1.2
R2D4_Slow cooling R2D4_Natural cooling R2D4_Water cooling

Relative initial tangent modulus

1.0

0.8

0.6 0.4

0.2

0.0 25 200 400 Temperature (C) 600 800

Figure 4.30. Relative initial tangent modulus vs. maximum temperature

Therefore the results show that faster cooling contributes to more significant decreases in both the strength and stiffness of concrete.

4.5.4. Comparison between Ep and Ei

The comparisons between elastic modulus ( E p ) obtained from the ultrasonic pulse velocity and initial tangent modulus ( Ei ) obtained from residual compression test are shown in Fig. 4.31 through Fig. 4.33. The trend for thermal degradation of E p and Ei is similar as shown in Fig. 4.31 through Fig 4.33. However, the elastic modulus ( E p ) is evaluated higher than the initial tangent modulus ( Ei ).

151

5.00E+06 4.50E+06 4.00E+06 3.50E+06 3.00E+06 E (psi) 2.50E+06 2.00E+06 1.50E+06 1.00E+06 5.00E+05 0.00E+00 25 200 400 Temperature (C) 600 800 Ep from UPV (Slow cooling) Ei_ initial tangent modulus (Slow cooling)

Figure 4.31. Comparison between E p and Ei (Slow cooling)

5.00E+06 4.50E+06 4.00E+06 3.50E+06 3.00E+06 E (psi) 2.50E+06 2.00E+06 1.50E+06 1.00E+06 5.00E+05 0.00E+00 25 200 400 Temperature (C) 600 800 Ep from UPV (Natural cooling) Ei_ initial tangent modulus (Natural cooling)

Figure 4.32. Comparison between E p and Ei (Natural cooling)

152
5.00E+06 4.50E+06 4.00E+06 3.50E+06 3.00E+06 E (psi) 2.50E+06 2.00E+06 1.50E+06 1.00E+06 5.00E+05 0.00E+00 25 200 400 Temperature (C) 600 800 Ep from UPV (Water cooling) Ei_ initial tangent modulus (Water cooling)

Figure 4.33. Comparison between E p and Ei (Water cooling)

4.6. Weight loss

Concrete is a porous material having many discrete and interconnected pores of different sizes and shapes. Water in the visible voids and large capillaries is called free water while the water in the combined form in hardened cement paste is called combined water or water of hydration [Ravindrarajah et al, (2000)]. When hardened concrete is gradually heated, the weight loss occurs. The weight loss relative to exposed temperature can be explained as follows. -Range 0-200 C (Drying stage): Evaporation of water from capillary pores -Range 200 C-600 C (Dehydration stage): Loss of non-evaporable water from gel pores can evaporate. The substantial shrinkage of concrete is accompanied.

153 The slope of weight loss is less than that in range 0-200C due to the difficulty in removing the water from smaller pore size. -Range 600 C-1000 C (Decomposition stage): The several changes including decomposition of CH (calcium hydroxide) and CHS (calcium silicate hydrate) phase in the cement paste occurs in the concrete system. - Above 1000 C: The combined water from hardened concrete is completely released. This causes the lost of the cementing property of concrete. The weight loss for the specimens with initial relative humidity 90 % and 100 % (fully saturated condition) are measured for each temperature history. The weight loss measurement does not include water cooling specimens because they are submerged in water immediately after the high temperature treatment.

Table 4.10. Effect of heating on weight loss (Initial relative humidity 90 %) Temp. (C) 25 200 400 600 800 R2D4S (%) 0 5.21 6.13 7.58 8.90 Initial relative humidity 90% R2D4N (%) R15D4N (%) 0 0 5.46 5.61 6.33 6.01 7.41 7.39 8.93 9.10

R2D2N (%) 0 5.03 6.04 7.22 8.85

Table 4.11. Effect of heating on weight loss (RH 100 %:Fully saturated) Temp. (C) 25 200 400 600 800 R2D4S (%) 0 8.03 9.33 9.69 10.04 Fully saturated R2D4N (%) R15D4N (%) 0 0 8.48 8.08 9.06 9.16 9.99 10.30 11.57 12.09

R2D2N (%) 0 7.78 8.91 9.82 11.00

154
14 12 10 Weight loss (%) 8 6
Notation R2D4S1(90%) R2D4N (90%) R15D4N (90%) R2D2N (90%) R2D4S1 (100%) R2D4N (100%) R15D4N (100%) R2D2N (100%)

4 2 0 0 100 200

R: heating rate D: holding time at max. temp. S1: cooling rate:1C/min N: natural cooling in the furnace

300

400

500

600

700

800

900

Temperature(C)

Figure 4.34. Effect of heating on weight loss (RH 100 % and 90 %)

Table 4.10 and 4.11 show the results of the weight losses of concrete specimens with initial RH 90 % and 100 %, respectively. The two initial RH values are used to study the effect of initial RH on weight loss of concrete under different temperatures. The weight losses for both RH 90 % and 100 % are plotted in Fig. 3.34. The weight loss in concrete increases as the temperature increases. However, the weight loss and maximum temperature is not linearly proportional over the entire temperature range. The slope of the weight loss plot above 200 C is considerably less than that in the range of 25-200 C because of different mechanisms of weight loss. In the lower temperature range, the weight loss is due mainly to the evaporation of free water in voids and capillary pores. When all free water is evaporated further weight loss comes from the removal of adsorbed water in smaller pores, which needs more energy. Moreover, at high temperature, some phase transformations take place and a certain amount of chemically bound water is released and evaporated. The

155 amount of adsorbed water and released chemically bound water is less than the free water in the concrete specimens with RH 90 % and 100 %. This leads to the reduced slope in the weight loss curve above 200 C. The initial weight loss plot of fully saturated specimens is steeper than those for RH 90 % specimens up to 200 C. This is because the fully saturated specimens contain more free water than the specimens of RH 90 %. At temperatures higher than 200 C, the slopes for both saturation levels are almost the same. This means that the evaporation of the free water in the voids and capillary pores is almost complete at a temperature of 200 C. Therefore, regardless of the initial moisture content of concrete, after 200 C, the rate of weight loss is similar for all specimens. This indicates that up to 800 C the weight loss of concrete is governed by the same mechanism. The test results, for both RH 90 % and RH 100 %, show that the weight loss is governed by target temperature, not by heating rate, cooling method, or holding time (2 hrs and 4 hrs in this research).

4.7. Cracks and color changes of the specimens

Red spots and distributed cracks were observed on every specimen to a maximum temperature of exposed to maximum temperature 600 C and above. It is likely that the distributed cracks on the bottom and top surfaces of the specimens occurred because of the differences in the thermal deformations between the aggregate and cement paste matrix. The chemical composition of granite is composed mostly of silica ( SiO2 ) and alumina ( Al2O3 ). The colors of the specimens subjected

156 to slow and natural cooling are generally pink, while the colors of the specimens subjected to water cooling are dark pink. The colors of the particles and aggregate on the specimens are gold in patches.

Cracks

(a) maximum temperature 600C

(b) maximum temperature 800C

Figure 4.35. Specimens exposed to heating rate 2 C/min (Natural cooling)

(a) Heating rate: 2C/min

(b) Heating rate: 15C/min

Figure 4.36. Specimens exposed to maximum temperature 800 C (Natural cooling)

The crack widths of the specimens exposed to a maximum temperature of 800 C are visibly wider than those of the specimens exposed to a maximum

157 temperature of 600 C in shown in Fig. 4.35. Also, in the specimens exposed to maximum temperature 800 C, the crack widths of the specimens heated quickly (15 C/min) are significantly larger than those of the specimens heated slowly (2 C/min) in shown in Fig. 4.36.

Figure 4.37. Granite concrete melted in the furnace

Generally, concrete is melted between about 1150 C and 1300 C. Fig. 4.37 shows a granite concrete lump (which was a 48 inch cylinder) inside the furnace melted from its exposure to the temperature of 1200 C.

4.8. Conclusions

1. There were three distinct peaks in the plot of the temperature difference between the surface and center of the specimen. The three peaks are related to the microstructural changes due to complex physicochemical transformations in the concrete

158 under high temperature. The first peak is associated with the dehydration of calcium silicate hydrate (C-S-H) and the evaporation of free water. The second peak is related to the decomposition of calcium hydroxide (CH) and calcium silicate hydrate (C-S-H). The third peak may be due to the decomposition of calcium carbonate ( CaCO3 ). 2. The results of the water permeability test (WPT) showed that the K p (coefficient of permeability) of the specimen increased as the temperature and cooling rate increases. Also, the test results of the specimens subjected to slow or natural cooling showed that the effect of the cooling rate on thermal damage in concrete is more significant than that of the holding time at target temperatures. 3. Strength and stiffness of the concrete decreased significantly as the maximum temperature and the cooling rate increase. 4. The trend for thermal degradation of E p (obtained from the ultrasonic pulse velocity) and Ei (obtained from residual compression test) was similar. However, the elastic modulus ( E p ) was evaluated higher than the initial tangent modulus ( Ei ). 5. The weight loss below 200 C was steeper than that above 200 C, for RH 90% and full saturation, because it is governed by the evaporation of the free water. The weight loss at temperatures above 200 C is due to the loss of adsorbed and chemically bound water in the specimen. The weight loss was governed by the target temperature rather than the heating rate, cooling method, or holding time (2 hrs and 4 hrs in this research). 6. The crack width increased with an increase in temperature and also with heating rate which means they both contribute to the damage of concrete.

159

CHAPTER 5

5.

A MULTISCALE CHEMO-MECHANICAL MODEL FOR MODULUS OF ELASTICITY OF CONCRETE UNDER HIGH TEMPERATURES

5.1. Introduction

Stiffness of concrete degrades under high temperatures. The degradation results mainly from two mechanisms. One is the variation of material properties of the constituent phases under high temperatures, and the other is the transformation of constituent phases under different temperatures. Therefore, the degradation of mechanical property of concrete under high temperatures must be studied from both mechanical and chemical points of view. Since the sizes of the constituent phases in concrete vary from centimeter to micrometer, and the phase transformations and property variations in the cement paste take place at a broad range of scale levels, a comprehensive mathematical model for characterizing the degradation of stiffness of concrete under high temperatures must be a multiscale model including the chemomechanical characteristics of the constituent phases under high temperatures. Extensive experimental studies have been performed by previous researchers. Piasta (1984 a) conducted experiments to study thermal deformation of phases

160 present in hardened cement paste and to determine initial temperatures in the range of 20-800 C which initiate the destruction of the micro structure of cement paste. Schneider and Herbst (1989) and Piasta and co-workers (1984 b) investigated chemical reactions and the behaviors of Ca(OH ) 2 , CaCO3 (calcite), C - S - H , nonevaporable water and micropores under various temperatures. In a study conducted by Lin et al. (1996) the microstructure of concrete exposed to elevated temperatures in both actual fire and laboratory conditions was evaluated with the use of ScanningElectron-Microscopy (SEM) and stereo microscopy. Also, Wang et al. (2005) used SEM to examine the cracking of high performance concrete (HPC) exposed to high temperatures under axial compressive loading of about 200 N. Despite the experimental studies, a prediction model has not yet been developed for the thermal degradation of stiffness of concrete in which phase transformations at the micro-scale level are considered. This study is a new attempt to predict the thermal degradation of the elastic modulus of concrete considering phase transformations under different temperature ranges. Stoichiometric models are used to handle the volumetric variations of the constituent phases, and composite mechanics models are used to obtain effective elastic modulus of concrete based on the volumetric variations under high temperatures. The degradation of stiffness of concrete results from the phase transformations taking place in the cement paste as well as in aggregates. In this study, we focused on phase transformations taking place in the cement paste, while the effect of various aggregates is included in the model by using a degradation factor representing

161 different aggregates. The internal structure of concrete can be divided into the following four scale levels (Constantinides and Ulm 2004): Level 1( 108 ~ 106 m ,the C - S - H level): A characteristic length scale of

108 106 m is the smallest material length scale. At this scale, the C - S - H exists in
at least two different forms with different volume fractions (inner and outer C-S-H). Level 2 ( 106 ~ 104 m , the cement paste level): Homogeneous C - S - H with large CH crystals, aluminates, cement clinker inclusions, and water. Level 3 ( 103 ~ 102 m , the mortar level): Sand particles embedded in a homogeneous cement paste matrix. Level 4 ( 102 ~ 101 m , the concrete level): Concrete as a composite material; course aggregates embedded in a homogeneous mortar matrix.

Level 4 (10 10 m )
-2 1

Level 3 (10 10 m )
-3 2

Level 2 (10 10 m )
-6 4

Level 1 (10-8 10 6 m )

Figure 5.1. Multiscale internal structure of concrete

In this study, the lowest scale level is at the micro-scale, i.e. the cement paste level (or Level 2). The decomposition of C-S-H under high temperatures will be considered in stoichiometric models, but the nanostructure of C-S-H will not be considered.

162 In Chapter 5.2, a kinetic model for hydration reactions of Portland cement will be described first. This model is necessary and important to calculate the initial volume fractions (prior to the exposure to high temperature) of the constituent phases in cement paste. In Chapter 5.3, initial volume fractions of the constituents in concrete will be calculated based on the kinetic model described in Chapter 5.2 and concrete mix design parameters. In Chapter 5.4, the variations of the initial volume fractions under elevated temperatures will be described by stoichiometric models based on phase transformations of the constituent phases taking place at different temperature ranges. In Chapter 5.5, the composite theory is briefly introduced which can be used at every scale level to obtain effective mechanical properties. In Chapter 5.6, the composite theory is used to predict the effective modulus of elasticity of concrete based on the volume fractions of the constituents at different temperature ranges. In Chapter 5.7, the model predictions are compared with available test data in the literature and with our own test data. Chapter 5.8 is conclusions.

5.2. Hydration kinetics model

Typically four clinkers ( C3 S , C2 S , C3 A , and C3 AF ) in Portland cement are considered as reactants in the hydration reactions of cement, which are considered to have five periods, namely the initial reaction period (Period 1), the induction period (Period 2), the acceleratory period (Period 3), the decelerating period (Period 4), and the slow period (Period 5). Periods 1 and 2 correspond to the early stage of hydration reaction, Periods 3 and 4 to the middle stage, and Period 5 to the late stage

163 (Taylor, 1997). The hydration reactions of cement are largely classified as two processes. The first one is the process of nucleation and growth reaction of the clinker phases, which is developed over 0-20 hours. The second is a diffusion controlled process where the kinetics of the hydration reaction is controlled by the rate of diffusion of dissolved ions through the layers of hydrates, which is formed around the clinker, and is developed at ages of 1 day and beyond. The rate of hydration of the second process depends strongly on the water to cement (w/c) ratio (Berliner et al, 1998). Fig. 5.2 is a schematic for the hydration process of cement paste.

Clinker phases Outer Product Clinker phases Outer Product Inner product Inner product

20hrs

After 20hrs

=1

water

Figure 5.2. Hydration process of cement paste

Eq. (5.1) (called Avrami equation) has been commonly employed to model the nucleation and growth reaction kinetics, the first 20-30% of the reaction, in cement chemistry:

ln[1 ( i oi )] = [k (t to ) m ]

(5.1)

In which, i is the hydration degree of reaction of clinker i at the time t . oi is the hydration degree of reaction of clinker i at the time to , when the nucleation and growth process dominate the hydration. k is a rate constant that considers the

164 effects of nucleation, multidimensional growth, geometric shape factors, and the diffusion process. m is an exponent defining the reaction order.

i = 1 exp[ai (t bi )c ]
i

(5.2)

In which, ai , bi , and ci are coefficients, which were determined empirically with the specific Portland cement. The empirical equation, Eq. (5.2) proposed by Taylor (1987), is based on the Avrami equation. The coefficients are shown in Table 5.1. Also, Eq. (5.2) is used as an approximate equation for the hydration reaction of cement paste for large values of time.

Table 5.1. Coefficients of ai , bi , and ci Compound C3 S C2 S C3 A C3 AF ai 0.25 0.46 0.28 0.26 bi 0.9 0.0 0.9 0.9 ci 0.7 0.12 0.77 0.55

For a long duration of time ( t > 1 day ), the hydration reaction, controlled by the rate of diffusion, has been addressed by researchers such as Berliner et al (1998), Fuji and Kondo (1974). According to Fuji and Kondo (1974), the rate of hydration reaction is

(1 i )1/ 3 = (2 Di )1/ 2 (t t * )1/ 2 / R + (1 i* )1/ 3

(5.3)

165 In which, i* is the degree of reaction of clinker i at the time t * . The

hydration reaction is governed by the rate of diffusion of dissolved ions. Di is the diffusion constant ( cm2 / h ) of clinker i , and R is the initial radius of the clinker grains. An average particle size 5 104 cm can be used as the initial radius of the clinker grains (R). The parameter Di and coefficients taken from Berliner (1998) and Bernard (2003) are summarized in Table 5.2.

Table 5.2. Diffusion constant and coefficients in the diffusion model Clinkers C3 S w/c 0.3 0.5 0.7 0.3 0.5 0.7 0.3 0.5 0.7 0.3 0.5 0.7 Diffusion model Di ( cm / h ) t * ( h)
2

i*

0.42 1010 2.64 1010 15.6 1010 6.64 1010 2.64 1010 0.42 1010 2.64 1010 15.6 1010 20 or 30 0.60

C2 S C3 A

C3 AF

The hydration reaction of the cement paste based on the diffusion theory is very fast as compared with that from the empirical equation, Eq. (5.2). The overall degree of hydration, , of the cement-based material systems is related to the individual degree of hydration of clinkers. The overall degree of hydration may be expressed with Eq. (5.4).

166

m (t ) (t ) = m
i i i i i

(5.4)

In which, mi = mC3S , mC2 S , mC3 A , and clinker phases in the cement.

mC4 AF are the mass fractions of the

5.3. Initial volume fractions of constituent phases in concrete

5.3.1. Volume fractions of the constituent phases at the cement paste level

The initial volume fraction of the phases is calculated on the basis of parameters and equations developed by Bernard (2003). At the cement paste level, the total volume is composed of reactants (remaining water and cement grains) and products of the hydration reactions (such as C - S - H , CH, products by aluminates, and capillary voids). The total volume is expressed with Eq. (5.5).

total Vlevel _ c. p = Vw (t ) + Vi ck (t ) + VC S H (t ) + VCH (t ) + VAL (t ) + Vcapillary voids (t ) i

(5.5)

The volume of remained water in the reactant phases is obtained by subtraction of the water consumed during hydration from the initial water content.

i Vw (t ) = Vwo Vw i (t ) 0 i

(5.6)

167 In which, Vw is the volume of remaining water, Vwo is the initial volume of
i water in the matrix, and Vw is the volume of the consumed water for complete i hydration of clinker i . Vw is calculated by Eq. (5.7).

i Vw * / M mi i = N w i i ; i* = oi = c o w / w Vc Vc mi i

(5.7)

In which, Vco is the initial cement volume, mi ( mC3 S , mC2 S , mC3 A , and mC4 AF ) is the mass fraction of clinker phases in the cement, and i is molar mass of phase i .
i N w = nw / ni denotes the number, nw , of moles of the consumed water during the

hydration of ni = 1 mol of the clinker phase i of mass density i* . For example, the hydration reactions of C3 S and C2 S compound are expressed by Eq. (5.8) and (5.9).

2C3S + 10.6 H C3.4 - S2 - H 8 + 2.6CH 2C2 S + 8.6 H C3.4 - S2 - H 8 + 0.6CH

(5.8) (5.9)

In Eq. (5.8), the ratio of consumed water to the hydration of C3 S is 5.3. Among them, 1.1 moles are chemically bound, and 2.9 moles are absorbed in the
C - S - H pores. In Eq. (5.9), the ratio of consumed water to the hydration of C2 S is
C C 4.3. Thus, N w 3S and N w 2 S are 5.3 and 4.3 respectively. Eq. (5.8) and (5.9) also show

168 that the hydration of C3 S and C2 S leads to the formation of 1.3 moles of CH and 0.3 moles of CH respectively. The volume of the hydrated clinker phases in the cement is calculated with Eq. (5.10). In Eq. (5.10), Vi ck _ o is the initial volume of the clinker phases in the cement.

Vi ck (t ) = Vi ck _ o [1 i (t )]

(5.10)

C - S - H and CH are produced by the hydration of C3 S and C2 S . The

volume of C - S - H is calculated with Eq. (5.11).

S S VC S H (t ) = VCC3S H C3 S (t ) + VCC2S H C2 S (t )

(5.11)

S S VCC3S H and VCC2S H , which are asymptotic volumes produced by the hydration

of C3 S and C2 S , are calculated by Eq. (5.12) and (5.13) respectively.

* S C3 S / C3S MC S mC3S VCC3S H C3 * = N C SS H ; C3 S = o3 = c o Vc Vc C S H / C S H mi

i

(5.12)

* S MC S mC2 S C2 S / C 2 S VCC2S H * C2 ; C2 S = o2 = c = N C SS H o Vc C S H / C S H Vc mi

(5.13)

169 The volume of CH is calculated using Eqs. (5.14) - (5.16), which are similar to Eqs. (5.11) - (5.13).

C3 C2 VCH (t ) = VCHS C3S (t ) + VCH S C2 S (t ) * C3 C S / C3S VCHS C3 = N CHS 3 CH / CH Vco * C2 C S / C2 S VCH S C2 = N CHS 2 CH / CH Vco

(5.14)

(5.15)

(5.16)

The parameters used in these equations are taken from Bernard (2003) and summarized in Table 5.3.

Table 5.3. Parameters for the determination of the volume fractions Parameters Reactants
C3 S C2 S C3 A C4 AF

Products
W C

C3.4 S 2 H 8

CH

i*[ g / cm3 ]
mi

* C3S

* C2 S

* C3 A

* C3 AF

1 18 -

3.15 -

2.04 227.2 -

2.24 74 -

0.543 228.32 1.0 1.3 5.3

0.187 172.24 1.0 0.3 4.3

0.076 270.20 10.0

0.073 430.12 10.75

i [ g / mol ]
N
i C S H i N CH i Nw

The capillary voids produced by the chemical shrinkage of the hydrates occurring during the hydration can be approximately calculated using Eq. (5.17) [see, Bentz (2006)].

170
Vcapillary voids = Cs c Vco (t )

(5.17)

Cs is the chemical shrinkage per gram of cement. The value of 0.07 ml / g by Bentz (2006) is used in this study. All volume fractions are calculated using
total Eq. (5.18). Vlevel _ c. p is calculated from the initial volume of the cement and water in

the mixture because the value is constant with time. The volume fraction occupied by aluminates is calculated from Eqs. (5.18) and (5.19).

fi =

i total level _ c . p

Vi w = Vi Vco 1 + c o w c V + Vw
o c

(5.18)

f AL = 1 f C S H + f i ck + f CH + f w + f capillary voids i

(5.19)

Under the condition of complete hydration ( = 1 ), the volume fractions of the cement paste, with a w/c ratio of 0.5 and 0.67, are calculated using the described equations and summarized in Table 5.4. The w/c ratio of 0.67 was used for residual compression tests in the present study.

Table 5.4. Volume fractions of constituents at cement paste level (w/c=0.5 and 0.67) Volume fractions fC S H fCH f AL f capillary void fw % (w/c=0.5) 53.69 15.71 15.76 8.56 6.28 % (w/c=0.67) 44.45 13.01 13.04 7.09 22.41

171 In the cement paste with w/c ratio of 0.5, the remaining water and capillary void form a macro porosity of 14.84 % ( f capillary void + f w = 14.84% ). These values agree with the results from Taylor (1997) and Hansen (1986) quite well.

5.3.2. Volume fractions of the phases at the mortar and concrete levels

The volume fractions of the constituent phases at the mortar and concrete levels are related to the mass proportions of the concrete mix design. At the mortar level, the volume fractions of the cement paste and sand can be calculated from Eq. (5.20).

fs =

f so us / s = ; f cp = 1 f s o o o fc + fw + f s uc / c + uw / w + us / s

(5.20)

In which, uc , uw , and us are the masses per unit volume for cement, water, and sand respectively. c , w , and s are the mass densities of cement, water, and sand respectively. f cp and f s are the volume fractions of cement paste and sand in mortar respectively. The volume fractions at the concrete level are obtained by considering the coarse aggregate (gravel) in Eq. (5.21).

fg =

f go f +f +f +f
o c o w o s o a

ug / g uc / c + uw / w + us / s + ug / g

; f m = 1 f g (5.21)

172 In which, ug is the mass of coarse aggregate per unit volume. g is the mass density of coarse aggregate. f m and f g are the volume fraction of mortar and coarse aggregate in concrete respectively.

Table 5.5. Density of different rock groups (Road research laboratory, 1959) Rock group Basalt Flint Granite Gritstone Hornfels Limestone Porphyry Quartzite Range of specific gravity 2.6-3.0 2.4-2.6 2.6-3.0 2.6-2.9 2.7-3.0 2.5-2.8 2.6-2.9 2.6-2.7 Average specific gravity 2.80 2.54 2.69 2.69 2.82 2.66 2.73 2.62 Density ( kg / m3 ) 2800 2540 2690 2690 2820 2660 2730 2620

Table 5.6. Volume fractions at mortar and concrete level (w/c=0.67) Contents Cement Sand Water Coarse agg Mix Design ( kN / m 3 ) 3.49 8.32 2.33 10.12 Density ( kg / m3 ) 3150.00 2690.00 1000.00 2690.00 fi o 0.012 0.032 0.024 0.039 Mortar level f s = 0.47 f cp =0.53 Concrete level f g =0.37 f m =0.63

5.4. A multi-scale stoichiometric model for phase transformations

5.4.1. Phase transformations at the cement paste level

It is difficult to calculate each phase transformation exactly with temperature increase because, with respect to temperature increase, the stoichiometric reactions

173 have not yet been conclusively established with experimental data. Therefore, some hypotheses are required to predict each phase transformation. In the present model, the following hypotheses are used. First, the total volume of hardened cement paste varies under different temperatures (Bazant and Kaplan, 1996). It expands during heating up to about 150 C; the maximum expansion is 0.2 %. No further expansion occurs between 150 C and 300 C. Between 300 C and 800 C the hardened paste shrinks, where shrinkage is between 1.6 and 2.2 % at 800 C. The expansion and shrinkage of cement paste are quite small comparing with the total volume. Therefore, the initial total volume of the cement paste is considered as a constant. Secondly, volume fraction of each phase in a phase transformation is assumed as a linear function between the beginning and ending temperature. For instance,
C - S - H is decomposed continually up to 800 C. The free water is evaporated at

about 100-120 C and bound water is gradually released up to 800 C. The loss of the bound water is assumed to be a linear function between 100 C and 800 C. Thirdly, loss of carbonation between 600C and 900C and the increase of calcite ( CaCO3 ) between 600 C and 800C are neglected. Lastly, the aluminum hydrates are regarded as non-reactive substances regardless of temperature increase. The capillary void, the remaining water, and the water in gel pores are regarded as the initial total void at the cement paste level.

f void _ cement = f w + f capillary void + f water in gel pores

(5.22)

174 By Copeland and Hayes (1953), the water volume in gel pores is about 28 % of the total volume of the gel. In the present model, the volume of the water in gel pores is considered as 28 % of C - S - H volume.

Table 5.7. Processes of decomposition depending on the temperature regime Temperature 20-120C 120-400C 400-530C 530-640C 640-800C Decomposition Evaporation of free water, Dehydration of C-S-H and ettringite Dehydration of C-S-H Dehydration of C-S-H, Dehydration of CH Dehydration of C-S-H, Decomposition of poorly crystallized CaCO3 Dehydration of C-S-H , Decomposition of CaCO3

Different processes of decomposition obtained from literature are summarized in Table 5.7 with respect to different temperature ranges (from 20 C to 800 C). Since the water in the gel pores is assumed as the initial void, it should be noted that Eq. (5.8) and (5.9) are balanced assuming all of the hydration products are saturated. After evaporation of free water in gel pores, the chemical formula of C-S-H is regarded as C3.4 - S2 - H 3 [Tennis and Jennings, 2000]. This means that 5 moles of H 2O (free water in gel pores) per 1 mole of C - S - H is evaporated between 100120 C. Thus, after evaporation of free water in gel pores, the decomposition of
C - S - H is described by Eq. (5.23). Between 400 C and 530 C, the decomposition

of calcium hydroxide (CH) is expressed using Eq. (5.24).

3.4CaO 2 SiO2 3H 2O 3.4CaO 2 SiO2 + 3H 2O Ca (OH ) 2 CaO + H 2O

(5.23) (5.24)

175

The volume of water decomposed from C - S - H and CH is regarded as an additive void. It is assumed that CH and C - S - H are completely decomposed at 530 C and 800 C, respectively. Under this assumption the volume fraction of the decomposed total water from C - S - H (between 120 C and 800 C) and CH (between 400 C and 530C) is calculated using Eq. (5.25).

i fi w = fi N w

i / i w / w

(5.25)

In which, fi w is the volume fraction of water decomposed from phase i [i= C3.4 S 2 H 3 and CH] in the cement paste. fi is the initial volume fraction of phase i. Particularly, it should be noted that fC S H is the volume fraction after evaporation of
i the free water in C-S-H gel pores. N w = nw / ni denotes the value nw as a number of
C CH moles of water decomposed from ni = 1 mol of phase i ( N w S H =3.0 and N w =1.0).

The density and molar mass of C3.4 S2 H 3 are 1.75 g / cm3 and 365 g / mol respectively (Tennis and Jennings, 2000). The parameters for CH and water are listed in Table 5.3. Finally, the volume of calcium oxide (CaO) produced from the decomposition of CH is obtained by deducting the decomposed water volume from initial volume of CH. From the same methodology, the volume of C3.4 S2 produced from the decomposition of C3.4 S2 H 3 is obtained by deducting the decomposed water volume from initial volume of C3.4 S 2 H 3 .

176

Table 5.8. Theoretical formulas for volume fraction change of each phase (w/c=0.67) Temperature
f C3.4 S2 H 3

Formulas (%) = 4.706 102 T + 37.649

120 C T 800 C

f C3.4 S2 = 3.488 102 T 4.185 f Cw S2 H 3 = 1.218 102 T 1.462 3.4

f CH = 0.10 T + 53.028 400 C T 530 C f CaO = 4.554 102 T 18.215

w f CH = 5.452 102 T 21.806

100

80 Relative volume (%)

Pore space

60

CH
40

CaO CS

C-S-H
20

Non-reactive substances
0 20 150 280 410 Temperature (C) 540 670 800

Figure 5.3. Change of phase composition with increasing temperature (w/c=0.67)

The theoretical formulas for the volume fraction change of each phase considering temperature ranges and w/c ratio are obtained from schemes described above. Table 5.8 shows the theoretical formulas of each phase in cement paste with a w/c ratio of 0.67 for the given temperature ranges. Fig. 5.3 shows the result for the

177 volume fraction change with the transformations of each phase up to 800 C using the formulas presented in Table 5.8.

5.4.2. Validation of current model at cement paste level

There has been no quantitative measurements considering changes of all the phases ( C - S - H , CH , CaO , solid grains, and so on) in cement paste with respect to temperature increase. Thus, the validation of the present model is done by (1) using the variation of pore volume under elevated temperatures measured experimentally by Piasta et al. (1984), and (2) comparing the theoretical calculation for the porosity change by Harmathy (1970).

Table 5.9. Mass fractions of clinker phases in the cement Clinker phases Piasta et al. (1984) Harmathy (1970) mC3 S 0.632 0.470 mC2 S 0.154 0.270 mC3 A 0.099 0.116 mC4 AF 0.080 0.090 Others 0.035 0.054

Table 5.9 shows the mass fractions of clinker phases for the cements used in the experiments of Piasta et al. (1984) and the theoretical calculation of Harmathy (1970). Piasta et al. (1984) performed various experimental studies related to the thermal properties of concrete with w/c of 0.4 at elevated temperatures. Harmathy (1970) calculated the porosity, the true density, and the bulk density of an idealized cement paste with w/c of 0.5 at elevated temperature using formulae based on the work of Powers (1960).

178 Table 5.10 shows the result of porosity tests between 20 C and 800 C by Piasta et al. (1984). To compute the volume fraction of the voids in cement paste from the test data of Piasta et al. (1984), the density of cement paste is approximated using Eq. (5.26).

cp =

mcp Vcp

c (1 + w / c ) mw + mc = Vw + Vc 1 + ( c / w ) ( w / c )

(5.26)

The volume fraction of the voids from the test data is calculated as the product of the density of cement paste using Eq. (5.26) and the value of total porosity.

Table. 5.10. Porosity and pore size distribution [w/c=0.4, (Piasta et al, 1984)]

In Eq. (5.6), Vw (t ) has a positive value beyond the w/c ratio of 0.45 for the cement used in their test. For w/c ratio of 0.4, the volume fraction of completely hydrated cement compounds is 0.89. The remaining 11 % of cement compounds exists as solid grains (non-reactive substances) in the cement paste. To reflect this

179 volume fraction of completely hydrated grains in Eqs (5.6), (5.11), (5.14), and (5.17),
i (t ) , the hydration degree, is multiplied by the factor of 0.89 for these equations. The

remaining procedures are the same as mentioned in Chapter 5.4.1.

100
w/c=0.4 [Piasta et al. (1984)] w/c=0.4 [Current model]

Volume fraction of pore space (%)

80

60

40

20

0 0 100 200 300 400 500 600 700 800 900

Temperature (C)

Figure 5.4. Comparison between current model and test data by Piasta et al. (1984)

100
w/c=0.5 [Harmathy (1970)] w/c=0.5 [Current model]

Volume fraction of pore space (%)

80

60

40

20

0 0 100 200 300 400 500 600 700 800 900

Temperature (C)

Figure 5.5. Comparison between current model and model of Harmathy (1970)

180

Table 5.11. Summary for the results shown in Fig. 5.4 and Fig. 5.5 Temp. (C) 100 200 300 400 500 600 700 800 Volume fraction of pore space (%) Prediction Harmathy Prediction Diff. (w/c=0.4) (w/c=0.5) (w/c=0.5) 23.28 0.65 44.00 28.04 24.43 1.99 47.00 29.18 25.86 2.05 51.00 30.61 27.29 0.95 52.00 32.05 35.58 6.90 54.00 39.11 39.07 2.10 57.00 42.23 40.51 7.30 58.00 43.66 41.94 1.58 61.00 45.10

Piasta et al. (w/c=0.4) 22.64 22.44 23.81 26.34 28.68 41.17 47.81 43.51

Diff. 15.96 17.82 20.39 19.95 14.89 14.77 14.34 15.90

The comparison for pore volume fractions between the present model and the test data by Piasta et al. (1984) is shown in Fig. 5.4. Fig. 5.5 shows the comparisons between the present model and the theoretical calculation by Harmathy (1970). Table 5.11 is a summary of the results shown in Fig. 5.4 and Fig. 5.5. The current model predicts the test results by Piasta et al. (1984) very well. The difference between the present model and the model by Harmathy (1970) is larger than the difference between the present model and the test results by the Piasta et al. because the model of Harmathy (1970) assumed that cement was only composed of C3 S (0.653 as the weight fraction) and C2 S (0.365 as the weight fraction). In other words, for the prediction of voids in a cement paste, Harmathys model only considered two phases ( C - S - H and CH), which are formed from the C3 S and C2 S components, as the total volume of cement paste. This assumption might cause a larger error than the present model because the model overestimates the volume of voids with respect to temperature increase.

181

5.4.3. Volume fractions of the phases according to temperature increase at the mortar and concrete level

At an elevated temperature, aggregates expand. However, the total volume of aggregates is assumed to be constant in the calculation for the volume fractions of sand and gravel.

Table 5.12. Volume fraction of phases at mortar level (w/c=0.67) Volume (%) C3.4 S2 H 3 C3.4 S2 CH CaO AL void _ cement sand Temperature (C) 20 C T 120 C 200C 400C 16.87 14.88 9.92 0.00 6.86 0.00 6.88 22.11 1.47 5.15 6.86 6.86 0.00 0.00 6.88 6.88 22.62 23.91 47.29

600C 4.96 8.82 0.00 3.12 6.88 28.93

800C 0.00 12.50 0.00 3.12 6.88 30.21

Table 5.13. Volume fraction of phases at concrete level (w/c=0.67) Volume (%) C3.4 S2 H 3 C3.4 S2 CH CaO AL void _ cement sand gravel Temperature (C) 400C 20 C T 120 C 200C 10.71 9.45 6.30 0.00 4.35 0.00 4.36 14.03 0.93 3.27 4.35 4.35 0.00 0.00 4.36 4.36 14.36 15.17 30.02 36.53

600C 3.15 5.60 0.00 1.98 4.36 18.36

800C 0.00 7.93 0.00 1.98 4.36 19.18

182 The volume fractions of each phase with respect to temperature increase at the mortar and concrete levels are calculated with Eq. (5.27) and (5.28).

f i '_ mortar = f i f cp _ mortar ; f s _ motar = f s , at mortar level f i '_ con = f i f cp f m ; f s _ con = f s f m ; f g _ con = f g , at concrete level

(5.27) (5.28)

In which, f i is the volume fraction for C3.4 S2 H 3 , CH , AL , C3.4 S2 , CaO , and void _ cement which changed with temperature increase at the cement paste level. Table 5.12 and Table 5.13 show volume fractions of each phase at the mortar level and concrete levels respectively (w/c=0.67).

5.5. Composite theories and damage theories

Based on composite mechanics theory the effective modulus, Eeff, of a twophase composite can be expressed with Eq. (5.29).

Eeff = f j ( c2 ) E1

(5.29)

In which, E1 is the modulus of phase 1 (may be considered as the matrix). c2 is the volume fraction of phase 2 (may be considered as the inclusion) for the total volume, which contains volume V1 for phase 1 and volume V2 for phase 2, in such a manner that V1 + V2 = V . Depending on the volume fraction and distribution of the

183 phase 2 in the phase 1, f j ( c2 ) represents the variation of the effective modulus due to the appearance of phase 2. Subscript j represents different types of distributions of the phase 2. If the phase 2 is distributed parallel to the loading direction, f j ( c2 ) is expressed with Eq. (5.30) and the corresponding Eeff gives the upper bound for the modulus.

f parellel ( c2 ) = (1 c2 ) + c2 ( E2 / E1 )

(5.30)

It is also known as Voigts model, an iso-strain model, because the strains in phase 1 and phase 2 are the same in a representative volume element (RVE). If the phase 1 and the phase 2 are distributed serial to the loading direction,
f j ( c2 ) is expressed with Eq. (5.31) and the corresponding Eeff gives the lower

bound for the modulus.

(1 c2 ) c2 + 1/ E2 E1 f serial ( c2 ) = E1

(5.31)

It is also known as Ruesss model, an iso-stress model, because the stresses in phase 1 and phase 2 are the same in RVE. If the phase 2 is in a spherical shape of different sizes and distributed randomly within phase 1, f j ( c2 ) is expressed with Eq. (5.32)

184 f spherical (c2 ) = 1 + c2 (1 c2 ) / 3 + 1 /[( E2 / E1 ) 1]

(5.32)

Eq. (32) is called the spherical model for damage distribution or generalized self-consistent model for effective modulus of two-phase composite. The corresponding Eeff is between the lower and upper bounds. As long as analytical solutions for f j ( c2 ) are available, Eq. (5.29) can be used for the effective modulus of composites with any distribution of phase 2, although only three different types of distributions, i.e. j = parallel, serial, and spherical are shown here. It is worthwhile to point out that when the parallel model is used and the phase 2 is considered as the damaged phase with modulus E2 = 0 , Eq. (5.29) and (5.30) become Eq. (5.33) and (5.34), respectively.

Eeff = (1 c2 ) E1

(5.33) (5.34)

f parellel (c2 ) = (1 c2 )

As one can see, Eq. (5.33) is the fundamental equation used in conventional scalar damage mechanics developed by Kachanov in 1958 (Lemaitre 1992; Krajcinovic 1996). This implies that the conventional scalar damage mechanics considered a special distribution for the damaged phase, i.e. parallel distribution of the damaged phase in distressed materials. Therefore, the conventional scalar damage

185 mechanics represents the upper bound of all possible damage distributions and developments. Recently developed composite damage mechanics is a more general damage theory, capable of handling various types of damage distributions and developments (Xi 2002; Xi and Nakhi 2005; Xi et al. 2006). In this study, the composite damage mechanics together with the spherical model is used, Eq. (5.29) and Eq. (5.32), under the assumption that the transformed phases in the concrete due to high temperatures are distributed randomly as spherical shape of different sizes within the original phases.

5.6. Thermal degradation of the modulus elasticity of concrete

When phase transformations take place in concrete under high temperatures. Some phases with high moduli convert to other phases with lower moduli and might even have a modulus equal to zero (i.e. void). Nucleation of voids in this case, is due to the evaporation of water at high temperature. As described in section 4, water includes the existing water in the concrete as well as the water generated by the phase transformations. By considering the phase transformations taking place in the concrete as independent processes (which could be simultaneous but with no coupling effect), the following equation, Eq. (5.35) for the effective modulus of concrete can be developed based on the composite damage mechanics, Eq. (5.29).

186

T , C3.4 S T T ,i T , CH Eeff = f spherical Econ _ ref = ( f spherical2 H 3 f spherical .... ) Econ _ ref

N T ,i = f spherical Econ _ ref i = C3.4 S2 H 3

(5.35)

Eq. (5.35) considers the phase transformations described in Chapter 5.4. The
T ,i function f spherical , which denotes the thermal degradation factor of the elastic modulus

from each original phase to their respective decomposed phases. Eq. (5.35) is a result of recursive applications of Eq. (5.29), with each transformation described by its own
T ,i function f spherical . For example, C3.4 S2 H 3 is decomposed into C3.4 S2 and H 2O

(considered as new void) with temperature increase as shown in Eq. (5.23). c2 is the volume fraction of the decomposed phase C3.4 S2 with respect to the original phase C3.4 S2 H 3 , and E2 / E1 is the ratio for the elastic modulus of the decomposed phase C3.4 S2 ( E2 ) to the original phase C3.4 S2 H 3 ( E1 ). The effective stiffness considering the original phase C3.4 S2 H 3 and the decomposed phase C3.4 S2 is obtained by putting Eq. (5.32) into Eq. (5.29). To obtain the finalized thermal degradation factor for the original phase C3.4 S2 H 3 , Eq. (5.32) is applied again for the second decomposed phase H 2O . In
C Eq. (5.33), the stiffness of original phase, E1 , is the effective stiffness Eeff
3.4 S2

_ C3.4 S2 H 3

which is obtained from C3.4 S2 H 3 and the previous transformation C3.4 S2 . The volume fraction and stiffness of the H 2O (new void) phase are c2 and E2 (where the phase is void, E2 =0). Eq. (5.36) is the function for the thermal degradation factor of the

187 elastic modulus transformed from C3.4 S2 H 3 (the original phase) to C3.4 S2 and H 2O (the respective decomposed phases).

T , C3.4 S f spherical2 H 3 = 1 +

cH 2O
C (1 cH 2O ) / 3 + 1/[( E H 2O / Eeff3.4 S2 _ C3.4 S2 H 3 ) 1]

(5.36)

In the determination of the order of the decomposed phases in application of the composite damage mechanics, it is noticed that Eeff obtained from the first decomposed phase should be higher value than the elastic modulus of the second decomposed phase because the thermal degradation factor should always be less than
T ,i or equal to 1. In summary, the modulus ratios in f spherical is not simply the stiffness

ratio of the product and the reactant, but the stiffness ratio of the product and the effective media. It is important to point out that Eq. (5.35) does not include the effect of thermal degradation of aggregates used in concrete. The aggregates experience various phase changes during thermal treatments of different heating and cooling rates. Therefore, another function, f agg_deg , for the thermal degradation of aggregates is introduced. Eq. (5.37) is the final thermal degradation function for the elastic modulus of concrete. Eq. (5.38) is the total thermal degradation factor of concrete

N T T ,i Eeff = f spherical f agg_deg Econ _ ref = F (T ) Econ _ ref i = C3.4 S2 H 3

(5.37)

188 N T ,i F (T ) = f spherical f agg_deg i = C3.4 S2 H 3

(5.38)

5.7. Comparison between present model and experimental results

To calculate the thermal degradation of elastic modulus of concrete using the present model, the stiffness of each phase must be evaluated. Table 5.14 is a summary for elastic properties of the constituent phases obtained from literatures.

Table 5.14. Elastic properties of constituent phases Phases

CSH

C3S C2 S

CH

CaO

C3 A C3 AF

E (Gpa ) 31 4 29.4 2.4 135 7 147 5 140 10 130 20 35.24 48 39.77 44.22 36 3 38 5 194.54 0.5 160 10 145 10 125 25

v ( ) 0.24 0.3 0.3 0.3 0.3 0.305 0.325 0.207 -

References Acker Constantinides and Ulm Acker Velez et al. Acker Velez et al. Beaudoin Wittmann Monteriro and Chang Acker Constantinides and Ulm Oda et al. Acker Velez et al Velez et al

It is noticed that the elastic modulus of CS decomposed from C - S - H and

CaO decomposed from CH vary depending on the porosity of the phases. The

porosities of CaO and C3.4 S2 are therefore calculated using the volume fractions of

189 the phases decomposed from CH and C3.4 S2 H 3 . The functions are shown in Eq. (5.39).

pCaO =

VH 2O _ CH VCaO _ CH + VH 2O _ CH

; pC3.4 S2 =

VH 2O _ C3.4 S2 H 3 VC3.4 S2 _ C3.4 S2 H 3 + VH 2O _ C3.4 S2 H 3

(5.39)

Table 5.15. Elastic moduli of each phase used in the present model Phases C3.4 S2 H 3 C3.4 S2
CH CaO Void

pi (Porosity) 0.26 0.54 -

E ( Gpa ) 32.0 29.79 [from E = 120 (1 pC3.4 S2 ) n , n=4.65] 40.2 8.35 [from E = 194.54 (1 pCaO )n , n=4] 0

Table 5.15 is a summary for the elastic modulus of each phase used in the present model. The functions for the elastic modulus of CaO and C3.4 S2 with respect to porosity are based on empirical functions by Velez et al (2001). The fixed variable (n) in the function, on Table 5.15, for the elastic modulus of CaO is assumed as 4, which is an average value for variables used in the functions for C2 S and C2 A by
T , AL Velez et al (2001). f spherical for the aluminum hydrates is 1 because they were

assumed to be a non-reactive substance. Thus, the elastic moduli of the aluminum hydrates are not used in the calculation of the thermal degradation.

f agg_deg = 0.03921 + e 0.002T

(5.40)

190 The thermal degradation ( f agg_deg ) of the aggregates is taken from Bazant and Kaplan (1996). Finally, the prediction model for the thermal degradation of elastic modulus can be expressed in terms of temperature, combining the relevant equations in previous sections. Since the capillary void, remaining water, and water in gel pores of
C - S - H are assumed as the initial total void from 20C to 120C in the model, the

thermal degradation factor of concrete for 20-120C is the same as the thermal degradation ( f agg_deg ) for the elastic modulus of the aggregates. Eq. (5.41) through (5.43) are the thermal degradation factors of concrete according to temperature ranges.

For 120-400C,
F (T ) = (39.21 103 + e 0.002T ) (697.126 103 253.828 106 T ) (651.437 103 + 126.914 106 T )

(5.41)

For 400-530C,
563.948 104 (3.921 102 + e 0.002T ) 2 5 3 6 (178.434 10 279.418 10 T ) (697.126 10 253.828 10 T ) F (T ) = (77.1825 + T ) (5132.91 + T )

(5.42) For 530-800C, F (T ) = 357.689 103 (3.921 102 + e 0.002T ) (697.126 103 253.828 106 T ) ( 651.437 103 + 126.914 106 T ) (5.43)

191 The prediction model is compared with the test data from previous researchers including our own study (Lee et al., 2006). Table 5.16 is a summary for the results of elastic modulus (chord modulus calculated by ASTM C 469) obtained from the residual compression test. Table 5.17 shows the values of the relative elastic modulus for the three cooling methods.

Table 5.16. Elastic moduli from residual compression test Temp.(C) 25 200 400 600 800 E : Elastic modulus (psi) Slow cooling Natural cooling Water cooling 2.96E+06 2.96E+06 2.96E+06 2.02E+06 1.94E+06 1.60E+06 1.07E+06 9.90E+05 4.53E+05 2.12E+05 1.64E+05 7.94E+04 4.61E+04 2.82E+04 1.87E+04 Table 5.17. Relative elastic modulus Temp. (C) 25 200 400 600 800 Relative elastic modulus Slow cooling Natural cooling Water cooling 1.0000 1.0000 1.0000 0.6818 0.6568 0.5414 0.3631 0.3345 0.1531 0.0715 0.0555 0.0268 0.0156 0.0095 0.0063

Fig. 5.6 shows the comparison between the present model and the test data in our study and by previous researchers. The present model satisfactorily predicts the trend of the test results. In the tests by previous researchers, the experimental conditions [such as w/c ratios, curing ages, heating rates, holding times at each target temperature, loading time (unstressed test and residual test)] are different from each other. However, it is shown in Fig. 5.6 that the trend for the thermal degradation of

192 the elastic modulus of concrete is similar. The model predicts the test results with relatively good accuracy.

1.2

1.0 Relative elastic modulus (E)

Slow cooling Natural cooing Water cooling Current model Pimienta and Hager (2002) Felicetti et al. (1996) Morita et al. (1992) Diederichs et al. (1988)

0.8

0.6

0.4

0.2

0.0 0 100 200 300 400 500 600 700 800 Temperature (C)

Figure 5.6. Comparison between current model and test data

5.8. Conclusions

1. A multiscale model was established for predicting the thermal degradation of the modulus of elasticity of concrete under different temperature ranges. The internal structure of concrete was considered at three scale levels: cement paste, mortar, and concrete. Cement paste is considered as a multiple phase composite with cement particles and all hydration products; mortar is a mixture of cement paste and sand; and concrete is a mixture of mortar and gravel.

193 2. At the cement paste level, the phase transformations in cement paste under high temperature were modeled by stoichiometric relations, which give the volume fractions of the products of the phase changes. The effective elastic modulus of cement paste under a certain temperature was then calculated based on composite models and using the volume fractions of the constituent phase at the temperature. The water released during the heating process is considered as part of voids with zero modulus of elasticity. 3. The variation of pore volume with increasing temperature was predicted by the present model, and compared with the test data by Piasta et al. (1984). The present model predicted the test results very well. The change of porosity by the present model was also compared with the model prediction by Harmathy (1970), which showed some difference. The difference was due to the assumption used in Harmathys model, that is, the cement is only composed of C3 S (0.653 as the weight fraction) and C2 S (0.365 as the weight fraction), and other phases were neglected. 4. At the mortar and concrete levels, the effective elastic moduli of mortar and concrete under a certain temperature was calculated based on composite models and using the volume fractions and moduli of cement paste, sand, and gravel. The elastic modulus of cement paste came from the effective modulus obtained from the lower level model, and the variations of elastic moduli of various sand and gravels were established based on available test data in the literature. 5. The present model for the thermal degradation of elastic modulus of concrete predicted satisfactorily the basic trend shown in the available test results which included various concrete mix design parameters and heating and cooling conditions.

194 6. The multi-scale chemo-mechanical model provided a general framework that can be used in the prediction for thermal degradations of elastic modulus as well as other properties of concrete, such as thermal strain and diffusivity of concrete under high temperatures.

195

CHAPTER 6

6.

A MULTISCALE CHEMO-MECHANICAL MODEL FOR THERMAL STRAIN OF CEMENT PASTE AND CONCRETE UNDER HIGH TEMPERATURES

6.1. Introduction

The expansion of concrete subjected to extreme elevated temperature is linked with intricate micro-structural variations. This study proposes a model to predict the thermal strains of cement paste and concrete considering micro-structural changes under elevated temperatures ranging from 20C to 800C. The model can consider characteristics of various aggregates in the calculation of thermal expansion for concrete. The model is a combination of the multi-scale stoichiometric model, which was described in Chapter 5.2 through 5.4, and a multi-scale composite model, which was proposed by Xi and Jennings (1997). At the cement paste level, the model satisfactorily predicted a test result. At the concrete level, upper bounds from the model were matched relatively well with test results by previous researcher. If the mechanical properties, such as elastic modulus (E), Poissons ratio (), and thermal deformation, of the aggregates used in concrete are given, it is likely that the model will reasonably predict experimental results. In Chapter 6.2, the model for shrinkage of cement paste and concrete using the effective homogeneous theory proposed by Xi

196 and Jennings (1997) is briefly introduced. In Chapter 6.3, the information for the stiffness, Poissons ratio, and thermal expansion of each phase are analyzed from literature review. The analyzed results are used in the model. In Chapter 6.4, the model predictions are compared with available test data in the literature. Chapter 6.5 is conclusions.

6.2. Model for shrinkage of cement paste and concrete

Concrete is a heterogeneous material in which constituents are distributed randomly. Thus, there is no exact solution in modeling material properties considering constituents in concrete. To simplify the internal structure of a material, composite models for effective properties have been developed. The three phase model developed by Cristensen (1979 a, 1979 b) was originally developed for elastic properties only, but Herve and Zauoi (1990) have shown that the model can be extended to nonlinear materials.

(a) Concrete

(b) Partitioning

(c) Spherical model

Figure 6.1. Simplification to spherical model

197

Phase1 Phase2 Effective homogeneous medium (Phase 3)

Figure 6.2. Three-phase effective media model

In Fig. 6.1, the meso-structure of concrete [Fig. 6.1 (a)] can be expressed as Fig. 6.1 (b) by partitioning aggregate and matrix. In the spherical model the partitions are simplified using spherical elements, Fig. 6.1 (c), such that the volume fraction of each phase and internal structure in an element are the same regardless of the size of elements. The spherical elements have three dimensional features, reducing the problem to one dimension [Xi and Jennings (1997)]. On the nanometer and micrometer scales, the spherical model can be applied. Fig. 6.2 shows the three-phase effective media model, where phase 3 is the effective homogeneous medium made equivalent to the heterogeneous medium. Xi and Jennings (1997) proposed a model for shrinkage of cement paste and concrete using the effective homogeneous theory. The proposed model is obtained by combining the model for shrinkage proposed by them and the multi-scale stoichiometric model described in Chapter 5. Although the model by Xi and Jennings is for shrinkage, it can also be used for effective thermal expansion in a heterogeneous medium. The effective bulk modulus and strain for the effective

198 homogeneous phase shown in Fig. 6.2 are expressed using Eq. (6.1) and (6.2) respectively.

12 K eff

V1 V + V ( K1 K 2 ) 1 2 = K2 + V1 K1 K 2 1 + 1 V1 + V2 K 2 + 4 G2 3

(6.1)

12 eff

V1 V1 K11 ( 3K 2 + 4G2 ) + K 2 2 1 V + V ( 3K1 + 4G2 ) V1 + V2 1 2 = V1 K 2 ( 3K 2 + 4G2 ) 4G2 ( K 2 K1 ) V1 + V2

(6.2)

In which, Ki , Vi , and Gi are the bulk modulus, volume fraction, and shear modulus of phase i , respectively. The general forms of effective bulk modulus and strain are delineated as Eq. (6.3) and (6.4), respectively [see, Xi and Jennings (1997)].

( Keff )i = Ki +

ci 1,i ( K eff )i 1 Ki ( K eff )i 1 Ki 1 + (1 ci 1,i ) 4 Ki + Gi 3

(6.3)

( eff )i =

( K ) ( )
eff i 1 eff

ci 1,i ( 3Ki + 4Gi ) + Ki i (1 ci 1,i ) 4Gi + 3 ( K eff )i 1 Ki 3 ( K eff )i 1 + 4Gi 4ci 1,iGi Ki ( K eff )i 1
i 1

(6.4)

In which, N i 2 ; ( K eff )1 = K1 and ( eff )1 = 1 . Parameter ci 1,i is:

199
ci 1,i = V j
j =1 i 1

V j for N > i 2 and cN 1,N = V j = 1 VN for i = N

j =1 j =1

N 1

(6.5)

The parameters Gi and Ki can be expressed easily in terms of elastic modulus and Poissons ratio. From elastic theory, Ki = Ei / 3(1 2 i ) and Gi = Ei / 2 (1 + i ) .

6.3. Material properties of constituents

To predict thermal expansion the information for the stiffness, Poissons ratio, and thermal expansion of each phase should be given. Experimental studies for heat deformations of cement paste were performed by Piasta (1984 a). Thermal strains of dehydrated and hydrated substances were measured using the dilatometer. The samples for the test were sleeve shaped with an inner diameter of 5 mm, an outer diameter of 10 mm, and a height of 50 mm. The dehydrated samples were compacted imparting a pressure of 40 Mpa. Paste samples were prepared with a w/c ratio of 0.5 and compacted by means of vibration. The investigation into heat deformation of the dehydrated substances was performed directly after forming. The paste samples were examined after curing for 28 days under a relative humidity of 95 % and a temperature of 20 2 C. A heating rate of 10 C/min was used to measure heat deformation for all substances. Thermal strain of CaO (decomposed from CH), was calculated using the test data from the coefficient of thermal expansion by Okayama (1978). Fig. 6.3 and Fig. 6.4 are the test data regarding thermal strains of dehydrated and hydrated substances respectively.

200

0.040

0.035

C2S
0.030

0.025

Strain

0.020

Potland Cement C4AF C3S CaO C3A

0.015

0.010

0.005

0.000 0 100 200 300 400 500 600 700 800 900

Temperature (C)

Figure 6.3. Thermal strain test data of dehydrated substances

0.020 Expansion 0.010

0.000

C2SH C4AFH Cement paste Ca(OH)2 C3SH C3AH

-0.010

Strain

Contraction

-0.020

-0.030

-0.040

Ettringit

-0.050 0 100 200 300 400 500 600 700 800 900

Temperature (C)

Figure 6.4. Thermal strain test data of hydrated substances

201 Generally up to 150 C hydrate substances expand because the expansion in the dehydrated parts is more prevalent than the shrinkage of hydrated parts [Piasta (1984 a)]. At temperatures above 150 C shrinkage prevails due to dehydration. In phases considered for the model at the cement paste level are C3.4 S2 H 3 ,

C3.4 S2 , CH , CaO , aluminum hydrates, and void space. The test data of CaO in
Fig. 6.3 and CH in Fig. 6.4 are used in the model. However, because there is no exact information for the thermal strain of the chemical components C3.4 S2 and

C3.4 S2 H 3 from literature, the thermal stain of C3.4 S2 is assumed on the basis of the
test data for C2 S and C2 S in Fig. 6.3. For C3.4 S2 H 3 , the average thermal strain of

C2 S H and C3S H is used in the model.

In the model, the thermal strain for alumina hydrates should be considered because the thermal deformations of alumina hydrates, i.e. ettringite and C3 AH , are large compared to other components as shown in Fig. 6.4. In cement paste, alumina hydrates exist in various chemical components. It is difficult to create a model that contains the all chemical components for alumina hydrates. Usually, the hydration of

C3 A and C3 AF in Portland cement involve reactions with sulfate ions which are
supplied by the dissolution of gypsum. The primary reactions of C3 A and C3 AF are expressed with Eq. (6.6) and (6.7) respectively. (See, Sidney et al, 2002).

C3 A + 3CSH 2 ( gypsum ) + 26 H C6 AS3 H 32 (ettringite) C4 AF + 3CSH 2 ( gypsum) + 21H C6 ( A, F ) S3 H 32 + ( A, F ) H 3

(6.6) (6.7)

202

Ettringite (Eq. 6.6) is the first hydrate to crystallize because of the high ratio of sulfate to aluminates in the solution phase during the first hour of hydration. In Portland cement, which contains 5-6 percent gypsum, ettringite contributes to early strength development. After the depletion of sulfate when the aluminate concentration goes up again due to renewed hydration of C3 A and C4 AF , ettringite becomes unstable and is gradually converted into monosulfate, the final product (Mehta, 1986). Eq. (6.8) contains the chemical reaction from ettringite to monosulfate. Eq. (6.9) shows the chemical reaction converting from C6 ( A, F ) S3 H 32 to 3 C4 ( A, F ) SH12 .

C6 AS3 H 32 ( ettringite) + 2C3 A + 4 H 3C4 ASH12 ( monosulfate) C6 ( A, F ) S3 H 32 + C4 AF + 7 H 3C4 ( A, F ) SH12 + ( A, F ) H 3

(6.8) (6.9)

In Eq. (6.7) and (6.9), iron oxide plays the same role as alumina during hydration. F can substitute for A in the hydration products. The use of a formula such as C6 ( A, F ) S3 H 32 indicates that iron oxide and alumina occur interchangeably in the compound, but the A/F ratio need not be the same as that of the parent compound. When monosulfate comes in contact with a new source of sulfate ions ettringite can be formed once again. This potential for reforming ettringite is the basis for sulfate attack on Portland cement (Sidney et al, 2002). In the present model alumina hydrates are assumed as monosulfate. However, the information for thermal strain of monosulfate could not be found in the literature.

203

Table 6.1. Summary for thermal strain of phases in cement paste used in model Phase Temperature 20 C < T 600 C 600 C < T 700 C 700 C < T 800 C 20 C < T 800 C 20 C < T 200 C 200 C < T 270 C 270 C < T 300 C 300 C < T 400 C 400 C < T 600 C 600 C < T 650 C 650 C < T 800 C 20 C < T 200 C 200 C < T 300 C 300 C < T 400 C 400 C < T 430 C 430 C < T 500 C 500 C < T 590 C 590 C < T 650 C 650 C < T 700 C 700 C < T 800 C 20 C < T 50 C 50 C < T 120 C 120 C < T 175 C 175 C < T 300 C 300 C < T 375 C 375 C < T 400 C 400 C < T 450 C 450 C < T 500 C 500 C < T 600 C 600 C < T 650 C 650 C < T 800 C Strain = 149.664 107 T 299.328 106 = 183.974 107 T 235.795 105 = 152.974 107 T 187.949 106 = 126.923 107 T 253.846 106 = 722.222 108 T 144.444 106 = 464.286 108 T + 222.286 105 = 178.846 107 T + 580.385 105 = 215.513 107 T + 690.385 105 = 321.667 107 T + 1.115 102 = 2.1 105 T + 4.45 103 = 5 106 T 1.245 102 = 111.111 107 T 222.222 106 = 2.2 105 T 2.4 103 = 8.0 106 T + 1.8 103 = 266.667 107 T 566.667 105 = 542.857 107 T + 291.429 104 = 144.444 106 T + 742.222 104 = 166.667 107 T 116.667 105 = 2.4 105 T + 3.6 103 = 8.0 106 T 7.6 103 = 2.2 105 T 4.4 104 = 8.0 106 T + 1.06 103 = 654.545 107 T + 795.455 105 = 1.44 104 T + 2.17 102 = 9.3 105 T + 6.4 103 = 1.13 104 T + 1.39 102 = 6.4 105 T 5.7 103 = 6.0 105 T 7.5 103 = 833.333 108 T 333.333 104 = 173.333 107 T 279.333 104 = 533.333 107 T 453.333 105

C3.4 S2
CaO

C3.4 S2 H 3

CH

3C4 ASH12 (Monosulfate)

204
0.020 Expansion

0.010

C3.4S CaO

0.000

-0.010

C3.4S2H3
Contraction

Ca(OH)2

Strain

-0.020

-0.030

-0.040

Monosulfate
-0.050

-0.060 0 100 200 300 400 500 600 700 800 900

Temperature (C)

Figure 6.5. Thermal strains of phases in cement paste used in model

As a matter of fact, there is no known mineral of monosulfate. Thus, the thermal strain of the monosulfate is assumed on the basis of the test data of C3 AH and ettringite shown in Fig. 6.4. Table 6.1 is a summary for thermal strain functions according to the temperature range of each phase in cement paste used in the model. Fig. 6.5 is a plot for the thermal strains of phases from the functions.

Table 6.2. Elastic modulus and Porosity of each phase Phases

pi (Porosity)
0.26 0.54 0.25 -

E ( Gpa )
32.0 29.79 [from E = 120 (1 pC3.4 S2 ) n , n=4.65] 40.2 8.35 [from E = 194.54 (1 pCaO )n , n=4] 40.0 0

C3.4 S2 H 3 C3.4 S2
CH CaO Monosulfate Void

205 Table 6.2 shows the porosity and elastic modulus of phases used in the model. The methodology to obtain the elastic modulus of each phase except for monosulfate was already described in Chapter 5.7. The elastic modulus and Poissons ratio for monosulfate are assumed as 4 Gpa and 0.25, respectively. Generally, the volume portion of aggregates is between 60 % and 80 % of the total volume of concrete. Therefore, they have a very important effect on the volume changes of concrete. The material properties of various aggregates are summarized in Table 6.3. It should be noticed that there is no consistency for the properties in the Table. 6.3 because the chemical portions constituting rock are different, even if the rocks are called by the same name.

Table 6.3. Elastic properties of various aggregates (Jumijis, 1983) Rock Basalt Diabase Gabbro Granite Syenite Dolomite Limestone Sandstone Shale (clay) Gneiss Marble Quartzite Schist Slate

Igneous rocks

Sedimentary rocks

Metamorphic

E (Gpa ) 19.1-111.5 22.0-114.0 58.4-107.8 21.3-68.5 58.8-86.3 19.6-93.0 8.0-78.5 4.9-84.5 7.8-44.0 14.2-70.0 28.0-100.0 25.5-97.5 4.0-70.5 -

v ( ) 0.14-0.25 0.103-0.333 0.125-0.48 0.125-0.338 0.15-0.319 0.08-0.37 0.10-0.33 0.066-0.62 0.04-0.54 0.091-0.25 0.11-0.38 0.11-0.23 0.01-0.20 0.06-0.44

Table 6.4 shows the thermal strain functions of sandstone and limestone obtained from curve fitting of test data by Soles and Geller (1964).

206 Table 6.4. Summary for thermal strain of limestone and sandstone Phase Sandstone Temperature 20 C < T 600 C Strain T 2.0 1011T 3 + 2.0 108 T 2
4

= 4.0 10

14

+ 9.0 106 T 0.0002

Limestone

600 C < T 650 C 650 C < T 800 C 20 C < T 800 C

= 1.75 105 T + 2.625 103 = 3.80 106 T + 1.153 102 = 4.0 1012 T 3 + 1.0 108 T 2 + 9.0 106 T 0.0003

6.4. Comparison between model and experimental results

To apply the multi-scale composite model, arrangement of the phases in cement paste is shown in Fig. 6.6.

Phase 1: Void Phase 2: Crystal Phases Phase1 Phase2 Phase3 Phase4 (CH & Monosulfate) Phase 3: Gel Phase (CSH) Phase 4: Cement paste (Effective medium)

Temperature increase (Decomposition of Phases) Phase 1: Void Phase 2: CaO Phase 3: CH Phase 4: Monosulafate Phase 5: CS Phase 6: CSH Phase 7: Cement paste (Effective medium)

1 2 3 4

6 7

Figure 6.6. Arrangement of phases in cement paste

207 Before cement paste is exposed to high temperature, void and crystal phases are surrounded by gel phase. When cement paste is exposed to high temperature, the decomposed solid phases, i.e. CaO from CH and C3.4 S2 from C3.4 S2 H 3 , are located in close proximity to primary phases and the decomposed voids from C3.4 S2 H 3 and
CH are located at the center of cement paste in the model. Finally, the model is

obtained by combination of the multi-scale composite model considering the arrangement for phases shown in Fig. 6.6 and the multi-scale stoichiometric model described in Chapter 5. The model is compared with the test result with w/c ratio of 0.5. The model satisfactorily predicts the test result as shown in Fig. 6.7.

0.005 0.003 0.001 -0.001 -0.003

Cement paste_Exp. [(Piasta, 1984a), w/c=0.5] Cement paste_Model (w/c=0.5)

Strain

-0.005 -0.007 -0.009 -0.011 -0.013 -0.015 0 100 200 300 400 500 600 700 800 900

Temperature (C)

Figure 6.7. Comparison between model and experimental data for cement paste

208 Table 6.5 is a summary for the thermal strain functions according to temperature rage of cement paste with w/c=0.5 obtained from the model.

Table 6.5. Thermal strain functions of cement paste from model (w/c=0.5) From model Temperature 20 C < T 50 C 50 C < T 120 C 120 C < T 175 C 175 C < T 300 C 300 C < T 400 C 400 C < T 425 C 425 C < T 500 C 500 C < T 530 C 530 C < T 575 C 575 C < T 625 C 625 C < T 650 C 650 C < T 800 C Strain = 117.667 107 T 235.333 106 = 462.857 108 T + 121.571 106 = 8.60 106 T + 1.709 103 = 30.032 106 T + 54.596 104 = 2.493 105 T + 3.929 103 = 3.80 105 T + 9.157 103 = 5.712 105 T + 17.283 103 = 3.66 105 T + 7.023 103 = 466.667 108 T 148.483 104 = 9.32 106 T 17.524 103 = 1.264 105 T 19.599 103 = 508.667 108 T 146.893 104

Cement paste

Phase 1: Aggregates Phase 2: Cement paste (From effective medium considering phases) Phase 3: Concrete Concrete (Effective homogeneous medium)

Phase1 Phase2

Figure 6.8. Arrangement of phases in concrete

In calculation of thermal strain of concrete from model, the arrangement of cement paste and aggregates is shown in Fig. 6.8. The aggregates are surrounded by

209 cement paste. The bulk moduli of cement paste according to temperature increase are calculated using Eq. (6.3) with the volume fractions of phases according to temperature increase from the multi-scale stoichiometric model. In application of Eq. (6.2) to calculate thermal expansion of concrete from the model, shear modulus of cement paste is calculated using the mixture theory of Eq. (6.10).

Gcp = f i Gi
i =1

(6.10)

In which, Gi and f i , which changed with temperature increase, are the shear modulus and the volume fraction for C3.4 S2 H 3 , CH , Monosulfate, C3.4 S2 , and CaO calculated at cement paste level.

0.020
LimestoneConc._Exp. [(Schneider, 1982), w/c=0.6] LimestoneConc._Upper bound from model (w/c=0.6) LimestoneConc._Lower bound from model (w/c=0.6)

0.015

0.010

Strain
0.005 0.000 -0.005 0 100 200 300 400 500 600 700 800 900

Temperature (C)

Figure 6.9. Comparison between model and experimental data (limestone concrete)

210

0.020
SandstoneConc._Exp. [(Schneider, 1982), w/c=0.62] SandstoneConc._Upper bound from model (w/c=0.62) SandstoneConc._Lower bound from model (w/c=0.62)

0.015

0.010

Strain
0.005 0.000 -0.005 0 100 200 300 400 500 600 700 800 900

Temperature (C)

Figure 6.10. Comparison between model and experimental data (sandstone concrete)

In the ranges of material properties of aggregates given in Table 6.3, the upper and lower bounds from the model for strains of limestone concrete (w/c = 0.6) and sandstone concrete (w/c = 0.62) are shown in Fig. 6.9 and Fig. 6.10 with the test data by Scheider (1982), respectively. The functions, which are based on the test data of Soles and Geller (1964), in Table. 6.4 were used as the thermal strains of the aggregates in the model. The test samples by Schneider (1982) were sleeve shaped with diameter 80 mm and height 300 mm. The samples were cured for 750 days under the condition of 65 % relative humidity at a temperature of 20 2 C after water curing for 7 days. The test data for the sandstone concrete is contained between the upper and lower bound from the model. While the test data of the limestone concrete is a little

211 higher than the upper bound from the model. The upper bounds from the model are matched relatively well with the test results. The rocks have different material properties due to different chemical portions in the rocks, even if they have the same name, the thermal strains are also different from each other. In the model, the test data of the aggregates by Soles and Geller (1964) was used to calculate the thermal strains of limestone concrete and sandstone concrete. The test data for the concretes used to compare with the model is from Schneider (1982). When aggregates and concrete (containing the same aggregates) undergo the same test conditions, and the material properties of the aggregates are given, it is likely that the proposed model predicts the thermal expansion of concrete reasonably.

6.5. Conclusions

1. At cement paste level, the model, obtained by a combination of the multi-scale composite model and the multi-scale stoichiometric model, was compared with test results from Piasta (1984 a), and satisfactorily predicted them. 2. The test data for the sandstone concrete was contained in the upper bound and lower bound from the model. While the test data for limestone concrete was a little higher than the upper bound from the model. 3. At concrete level, the upper bounds from the model matched relatively well with test results by Scheider (1982). 4. The thermal deformation of the aggregates used is an important factor in the thermal deformation of concrete because aggregates occupy about 70 % of the total

212 volume of concrete. When aggregates and concrete (containing the same aggregates) undergo the same test conditions, and the material properties of the aggregates are given, it is likely that the proposed model will predict the thermal expansion of concrete reasonably.

213

CHAPTER 7

7.

SUMMARY AND CONCLUSIONS

The studies performed in this thesis are largely classified as two experimental studies and two theoretical models. One of the experimental studies is to find the effects of temperature and moisture on strain of concrete and the other one is to investigate the strength, stiffness, and durability performance of concrete subjected to various heating and cooling scenarios. In Chapter 3, the effects of temperature and moisture on strain of concrete were investigated. The strain caused by temperature increase without moisture control (Conventional Thermal Strain: CTS), the strain caused by temperature increase under constant humidity (Pure Thermal Strain: PTS), and the strain caused by moisture change under constant temperature (Pure Hygro Strain: PHS) were measured continually and simultaneously over time. From the measured strains, the thermohygro coupling effect in the temperature range 28-70 C was obtained. The conclusions from the study are as follows; - If it is assumed that a local thermodynamic equilibrium always exist between the phases of pore water (vapor, liquid) within a very small element of concrete, the phenomena, which is the increase of internal relative humidity of concrete with

214 temperature increase, can be explained using Eq. (3.12). This equation is based on static force equilibrium from the capillary tube model and physicochemical equilibrium by the Kelvin equation. When we consider the strain change of aggregates and hardened cement paste as components of concrete at elevated temperatures, it is clear that the internal relative humidity of concrete increases as the temperature increases. - Pure Thermal Strain (PTS), because the there is no additional hygro strain by swelling pressure increase (increase of relative humidity) according to temperature increase, was less than Conventional Thermal Strain (CTS) in the range of temperatures used in this study. - Pure Hygro-Strain (PHS) was nonlinear. At higher temperatures, Pure Hygro-Strain (PHS) increased more rapidly. It is likely linked to complicated physiochemical reactions in concrete as humidity changes, i.e. the increase of swelling pressure (RH increase) is not directly proportional to temperature increase. - The thermo-hygro coupling effect was negative in the tested temperature range. That means that the internal relative humidity of concrete increases as the temperature increases in the tested temperature range. - The shrinkage effect due to a decrease in internal relative humidity (dehydration) of cement may be contained in the overall expansion of concrete beyond 150 C. However, it should be noticed that concrete expands as temperature increases because aggregates, which make up about 70 % of concrete by volume, expand continually as the temperature increases.

215 In Chapter 4, the strength, stiffness, and durability performance of concrete subjected to various heating and cooling scenarios were investigated. The strength and stiffness performance of concrete were investigated with the aid of Ultrasonic Pulse Velocity test (UPV) and residual compressive strength testing. The durability of the concrete was investigated using a water permeability test (WPT). Additionally, the thermal diffusivity, weight losses, color changes, and cracks of the specimens were reported. The conclusions from the study are as follows; - There were three distinct peaks in the plot of the temperature difference between the surface and center of the specimen. The three peaks are related to the micro-structural changes due to complex physicochemical transformations in the concrete under high temperature. The first peak is associated with the dehydration of calcium silicate hydrate (C-S-H) and the evaporation of free water. The second peak is related to the decomposition of calcium hydroxide (CH) and calcium silicate hydrate (C-S-H). The third peak may be due to the decomposition of calcium carbonate ( CaCO3 ). - The results of the water permeability test (WPT) showed that the K p (coefficient of permeability) of the specimen increased as the temperature and cooling rate increases. Also, the test results of the specimens subjected to slow or natural cooling showed that the effect of the cooling rate on thermal damage in concrete is more significant than that of the holding time at target temperatures. - Strength and stiffness of the concrete decreased significantly as the maximum temperature and the cooling rate increase.

216 - The trend for thermal degradation of E p (obtained from the ultrasonic pulse velocity) and Ei (obtained from residual compression test) was similar. However, the elastic modulus ( E p ) was evaluated higher than the initial tangent modulus ( Ei ). - The weight loss below 200 C was steeper than that above 200 C, for RH 90 % and full saturation, because it is governed by the evaporation of the free water. The weight loss at temperatures above 200 C is due to the loss of adsorbed and chemically bound water in the specimen. The weight loss was governed by the target temperature rather than the heating rate, cooling method, or holding time (2 hrs and 4 hrs in this research). - The crack width increased with an increase in temperature and also with heating rate which means they both contribute to the damage of concrete. Variations of concrete under high temperatures result mainly from two mechanisms. One is the variation of material properties of the constituent phases under high temperatures, and the other is the transformation of constituent phases under different temperatures. Therefore, the properties of concrete under high temperatures must be studied from both mechanical and chemical points of view. In Chapter 5, the model for the thermal degradation of elastic modulus of concrete was proposed by composite mechanics at three scale levels: concrete, mortar, and cement paste level. At the level of cement paste, the variations of volume fractions of the constituents were evaluated based on phase transformations taking place under different temperature ranges. Stoichiometric models were used to calculate the volumetric changes of the constituents. The conclusions from the study are as follows;

217 - A multiscale model was established for predicting the thermal degradation of the modulus of elasticity of concrete under different temperature ranges. The internal structure of concrete was considered at three scale levels: cement paste, mortar, and concrete. Cement paste is considered as a multiple phase composite with cement particles and all hydration products; mortar is a mixture of cement paste and sand; and concrete is a mixture of mortar and gravel. - At the cement paste level, the phase transformations in cement paste under high temperature were modeled by stoichiometric relations, which give the volume fractions of the products of the phase changes. The effective elastic modulus of cement paste under a certain temperature was then calculated based on composite models and using the volume fractions of the constituent phase at the temperature. The water released during the heating process is considered as part of voids with zero modulus of elasticity. - The variation of pore volume with increasing temperature was predicted by the present model, and compared with the test data by Piasta et al. (1984). The present model predicted the test results very well. The change of porosity by the present model was also compared with the model prediction by Harmathy (1970), which showed some difference. The difference was due to the assumption used in Harmathys model, that is, the cement is only composed of C3 S (0.653 as the weight fraction) and C2 S (0.365 as the weight fraction), and other phases were neglected. - At the mortar and concrete levels, the effective elastic moduli of mortar and concrete under a certain temperature was calculated based on composite models and using the volume fractions and moduli of cement paste, sand, and gravel. The elastic

218 modulus of cement paste came from the effective modulus obtained from the lower level model, and the variations of elastic moduli of various sand and gravels were established based on available test data in the literature. - The present model for the thermal degradation of elastic modulus of concrete predicted satisfactorily the basic trend shown in the available test results which included various concrete mix design parameters and heating and cooling conditions. - The multi-scale chemo-mechanical model provided a general framework that can be used in the prediction for thermal degradations of elastic modulus as well as other properties of concrete, such as thermal strain and diffusivity of concrete under high temperatures. In Chapter 6, the model for the thermal strain of cement paste and concrete, considering micro-structural changes under elevated temperatures ranging from 20 C to 800 C, was proposed. The model was obtained by the combination of a multiscale stoichiometric model and a multi-scale composite model. The model can consider characteristics of various aggregates in the calculation of thermal expansion for concrete. The conclusions from the study are as follows; - At cement paste level, the model, obtained by a combination of the multiscale composite model and the multi-scale stoichiometric model, was compared with test results from Piasta (1984 a), and satisfactorily predicted them. - The test data for the sandstone concrete was contained in the upper bound and lower bound from the model. While the test data for limestone concrete was a little higher than the upper bound from the model.

219 - At concrete level, the upper bounds from the model matched relatively well with test results by Scheider (1982). - The thermal deformation of the aggregates used is an important factor in the thermal deformation of concrete because aggregates occupy about 70 % of the total volume of concrete. When aggregates and concrete (containing the same aggregates) undergo the same test conditions, and the material properties of the aggregates are given, it is likely that the proposed model will predict the thermal expansion of concrete reasonably.

220

8.

REFERNCES

[1]

Abrams, M.S. (1971). Compressive strength of concrete at temperatures to 1600F, American Concrete Institute (ACI) SP 25, Temperature and Concrete, Detroit, Michigan.

[2]

Acker, P. (2001). Micromechanical Analysis of Creep and Shrinkage Mechanisms, Creep, Shrinkage and Durability Mechanics of Concrete and Other Quasi-Brittle Materials, Proc. of the 6th International Conference CONCREEP6, Elsevier, Oxford, UK, 15-25.

[3]

American Concrete Institute 216. (1981). Guide for Determining the Fire Endurance of Concrete Elements, Concrete International, Feb., 13-47.

[4]

American Concrete Institute 318. (2002). Building Code Requirements for Structural Concrete, American Concrete Institute

[5]

Anderberg, Y., and Thelandersson, S. (1976). Stress and deformation characteristics of concrete at high temperatures, Bulletin 54, Lund Institute of Technology, Sweden.

[6]

ASCE. (1992). Structural Fire Protection, ASCE Manuals and Reports on Engineering Practice, No.78.

[7]

Baant, Z., and Kaplan, M.F. (1996). Concrete at high temperatures, material

properties, and mathematical models, Longmon, Burnt Mill, U.K.

221 [8] Baant, Z., and Thonguthai W. (1979). Pore pressure in heated concrete walls theoretical prediction, Magazine of Concrete Research, 31(107), 67-75. [9] Beaudoin, J.J. (1983). Comparison of mechanical properties of compacted calcium hydroxide and Portland cement paste systems, Cement and Concrete Research, 13, 319-324. [10] Benz, D.P. (2006). Influence of water-to-cement ratio on hydration kinetics: Simple models based on spatial considerations, Cement and Concrete Research, 36(2), 238-244. [11] Berliner, R., Popovici. M., Herwig, K.W., Berliner, M., Jennings, H.M., and Thomas, J.J. (1998). Quasielastic neutron scattering study of the effect of water-to-cement ratio on the hydration kinetics of tricalcium silicate, Cement and Concrete Research, 28(2), 231-243. [12] Bernard, O., Ulm, F.J, and Lemarchand, E. (2003). A multiscale micromechanics-hydration model for the early-age elastic properties of cement-based materials, Cement and Concrete Research, 33, 1293-1309. [13] Bilodeau, A., Malhotra, V.M., and Hoff, G.C. (1998). Hydrocarbon fire resistance of high strength normal weight and light weight concrete incorporating polypropylene fibers, International Symposium on High Performance and Reactive Powder Concretes, Sherbrooke, QC, 271-296. [14] Castillo, C., and Durani, A.J. (1990). Effect of transient high temperature on high-strength concrete, ACI Material Journal, Jan/Feb, 87(1), 38-67. [15] Cheng, F.P., Kodur, V.K.R., and Wang, T.C. (2004). Stress-strain curves for high strength concrete at elevated temperature, Journal of Materials in Civil Engrg., ASCE, Jan/Feb., 16(1), 84-90.

222

[16]

Christensen, R. M. (1979 a). Mechanics of composite materials, J. Wiley, 2nd edition.

[17]

Christensen, R. M. and Lo, K. H. (1979 b). Solutions for effective shear properties in three phase sphere and cylinder models, Journal of Mech. Phys. Solids. 27, 315-330.

[18]

Consolazio, G.R., and Chung, J.H. (2004). Numerical simulation of nearsurface moisture migration and stress development in concrete exposed to fire, Computers and Concrete, 1(1), 31-46.

[19]

Constantinides, G. and Ulm, F.J. (2004). The effect of two types of C-S-H on the elasticity of cement-based materials: Results from nanoindentation and micromechanical modeling, Cement and Concrete Research, 33, 1293-1309.

[20]

Copeland, L.E. and Bragg, R.H. (1953). The determination of non-evaporable water in hardened Portland cement paste, ASTM Bulletin No. 194. ASTM, Philadelphia, 70-74.

[21]

Crowley, M.S. (1956). Initial thermal expansion characteristics of insulating refractory concretes, Bulletin of the American Ceramic Society, 35(12), 465468.

[22]

Cruz, C.R. and Gillen, M. (1980). Thermal Expansion of Portland Cement Paste, Mortar, and Concrete at High Temperatures, Fire and Materials, 4(2), 66-70.

[23]

David, L (2004). Experimental investigation of concrete subjected to fire loading: micromechanical approach for determination of the permeability, Masters thesis, Univ. of Vienna, Austria.

223

[24]

Dettling, H. (1964). The thermal expansion of hardened cement paste, aggregates, and concrete (in German). Deutscher Ausschuss fr Stahlbeton, Bulletin No. 164, W. Ernst & Sohn, Berlin, 1-64 (translation No. 458, PCA Technical Information Department, Cicago)

[25]

Diamond, S. (1976). Cement paste micro structure: an overview at several levels in hydraulic cement pastes- Their structures and properties, Conference, University of Sheffield, UK, Apr. pp.334.

[26]

Dias, W.P.S., Khoury, G.A., and Sullivan, P.J.E. (1990). Mechanical properties of hardened cement paste exposed to Temperatures up to 700 C (1292 F). ACI Materials Journal, 87(2), 160-165.

[27]

Diederichs, U., Jumppanen, U.M., Schneider, U. (1995). High temperature properties and spalling behavior of high strength concrete, Proceedings of the Fourth Weimar Workshop on High Performance Concrete: Material Properties and Design, Hochschule fuer Architektur und Bauwesen (HAB), Weimar, Germany, Oct. 1995, 219-236.

[28]

di Prisco, M, Felicetti, R., and Gambarova, P.G. (2003). On the fire behavior of SFRC and SFRC structures in tension and bending, 4th Int. Workshop on High-Performance Fiber Reinforced Cement Composites HPFRCC-4, Ann Arbor (Michigan, USA), June, 14.

[29]

Diederichs, U., Jumppanen, U-M, and Penttala, V. (1988). Material properties of high strength concrete at elevated temperatures, IABSE 13TH Congress, Helsinki, June.

[30]

Dotreppe, J.C., Franssen, J.M., Bruls, A., Baus, R., Vandevelde, P., Minne, R., Nieuwenburg, D., and Lambotte, H. (1996). Experimental research on the

224 determination of reinforced concrete columns under fire conditions, Magazine of Concrete Research, June, 49(179), 117-127. [31] ENV 1992-1-2, (1995). Design of concrete structures-Part 1-2: General rulesstructural fire design, European Committee for Standardization, Brussels. [32] ENV 1993-1-2, (1995). Design of steel structures-Part 1-2: General rulesstructural fire design, European Committee for Standardization, Brussels. [33] Eskandari-Ghadi, M., Xi, Y., and Sture, S. (2005). Cross-property relations between mechanical and transport properties of composite materials, submitted to Composites Part B: Engineering. [34] Felicetti, R., Gambarova, P.G., Rosati, G.P., Corsi, F., and Giannuzzi, G. (1996). Residual mechanical properties of high-strength concrete subjected to high-temperature cycles, Proceedings, 4TH International Symposium on Utilization of High-Strength/High-Performance Concrete, Paris, France, 579588. [35] Felicetti, R., and Gambarova, P.G. (1999). Residual capacity of HSC thermally damaged beams, Journal of Structural Engineering, ASCE, March, 125(3), 319-327. [36] Felicetti, R., and Gambarova, P.G. (2003). Heat in concrete: special issues in materials testing, Studies and Researches, Vol. 24, Graduate School in Concrete Structures-Fratelli Pesenti Politecnico di Milano, Italy. [37] Felicetti, R., Gambarova, P.G., and Meda, A. (2004). Guidelines for the structural Design of Concrete Buildings Exposed to Fire, Workshop fib Task Group 4.3.2, Fire Design of Concrete Structures, Politecnico di Milano, Italy, Dec.2-4, 2004.

225

[38] [39]

Fintel, M. (1985). Handbook of Concrete Engineering, 2nd edition. Foster, S. J., Burgess, L.W., and Plank, R.J. (2004). High-Temperature Experiments on Model-Scale Concrete Slabs at High Displacement, Structure and Fire -Third International Workshop, Ottawa, May 2004.

[40]

Fuji, K., and Kondo, W. (1974). Kinetics of the hydration of tricalcium silicate, Journal of the American Ceramic Society, 57(12), 492-502.

[41]

Furumura, F., Abe, T., and Shinohara, Y. (1995). Mechanical properties of high strength concrete at high temperature, Proceedings of the Fourth Weimar Workshop on High Performance Concrete: Material Properties and Design, Hochschule fuer Architektur und Bauwesen (HAB), Weimar, Germany, Oct., 237-254.

[42]

Grasley, Z.C. and D.A. Lange. (2004). Thermal Dilation and Internal Relative Humidity in Hardened Cement Paste, on CD-ROM proceedings of the ACBM/RILEM International Symposium on Advances in Concrete through Science and Engineering, Evanston, IL, March 21-24.

[43]

Grosshandler, W.L. (2002). Fire Resistance Determination and Performance Prediction, Research Needs Workshop: Proceedings, National Institute of Standards and Technology, NISTIR 6890, Sep.

[44]

Guo, J.S., and Waldron, P. (2000). Development of the stiffness damage test (SDT) for characterization of thermally loaded concrete, Materials and Structures, Oct., 33, 483-491.

[45]

Hansen, P.A., and Jensen, J.J. (1995). High Strength Concrete, Phase 3, Fire Resistance and Spalling Behavior of LWA Beams, SP6-Fire Resistance,

226 Report 6.3, SINTEF NBL-Norwegian Fire Research Laboratory, STF25 A95004, March. [46] Hansen, T.C. (1986). Physical structure of hardened cement paste. A classical approach, Material and Structures, 19(114), 423-436. [47] Harada, T., Takeda, J., Yamane, S, and Furumura, F. (1972). Strength, elasticity and thermal properties of concrete subjected to elevated temperatures, In international Seminar on Concrete for Nuclear Reactors, ACI Special publication No. 34, Vol.1, Paper SP34-21. ACI, Detroit, 377-406. [48] Harmathy, T.Z. (1970). Thermal Properties of Concrete at Elevated Temperature, ASTM Journal of Materials, 5(1), March, 47-74. [49] Harmathy, T.Z. (1993). Fire Safety Design & Concrete, Longman Scientific & Technical. [50] Herve, E. and Zaoui, A. (1990). Modeling the effective behavior of nonlinear matrix-inclusion composites, Eur. J. Mech., 9(6), 505-515 [51] Holmes, M., Anchor, R.D., Cook, G.M.E., and Crook, R.N. (1982). The effects of elevated temperatures on the strength properties of reinforcing and prestressing steels, Structural Engineer, March, 60B(1), 7-13. [52] Illston, J.M., and Sanders, P.D. (1973). The effect of temperature change upon the creep of concrete of torsional loading, Magazine of Concrete Research, 37(132), 131-144. [53] Joshi, R.C., Chatterji, S., Achari, G., and Mackie, P. (1999). Technical note : re-examination of ASTM C1202, September.

227 [54] Jumijis, A.R. (1983), Rock Mechanics, Trans tech publications, Houston, Gulf Publication, Co. [55] Khoury, G.A, Sullivan, P.J.E., and Grainger, B.N. (1984). Radial temperature distributions within solid concrete cylinders under transient thermal states, Magazine of Concrete Research, 36(128), 146-156. [56] Khoury, G.A, Sullivan, P.J.E., and Grainger, B.N. (1985a). Transient thermal strain of concrete: literature review, conditions within specimen and individual constituent behavior, Magazine of Concrete Research, 37(132), 131-144. [57] Khoury, G.A., Sullivan, P.J.E., and Grainger, B.N. (1985b). Strain of concrete during first heating to 600C under load, Magazine of Concrete Research, 37(133), 195-215. [58] Khoury, G.A., Grainger, B.N., and Sullivan, P.J.E. (1986). Strain of concrete during first cooling to 600C under load, Magazine of Concrete Research, 38(134), 3-12. [59] Khoury, G.A. (2000). Effect of fire on concrete and concrete structures, Prog. Struct. Engng Materials, Part 2, 429-447. [60] Khoury, G.A., Majorana, C.E., Pesavento, F., and Schrefler, B.A. (2002). Modelling of heated concrete, Magazine of Concrete Research, 54(2), 77101. [61] Kodur, V.K.R., and Sultan, M.A. (1998). Structural behavior of high strength concrete columns exposed to fire, International Symposium on High Performance and Reactive Powder Concretes, Sherbrooke, Canada, 217-232.

228 [62] Kodur, V.K.R. (1999). Fire performance of high-strength concrete structural member, Institute for research in construction, National Research Council of Canada, Ottawa, Ontario, Construction Technology Update No. 31. [63] Kodur, V.K.R. (2000). Spalling in high strength concrete exposed to fireconcerns, cause, critical parameters and cures, Proceedings, ASCE Structures Congress, Philadelphia, PA. [64] Kodur, V.K.R., and Sultan, M.A. (2003). Effect of temperature on thermal properties of high-strength concrete, Journal of Materials in Civil Engrg, ASCE, March/April, 15(2), 101-107. [65] Komarovskii, A.N. (1965). Design of nuclear plants, 2nd edition. Atomizdat, Moscow, Chapter 7, (translated from Russian by Israel Program for Scientific Translations, Jerusalem, 1968). [66] Komlo, K., Popovics, S., Nrnbergerov, T., Babl, B., and Popovics, J.S. (1996). Ultasonic pulse velocity test of concrete properties as specified in various standard, Cement and concrete composition, 18, 357-364. [67] Krajcinovic, D. (1996). Damage mechanics, North-Holland, Amsterdam, Netherlands. [68] Lach, V. (1970). ber die rehydratation von Portlandzement, Zement-KalkGips, No. 2. [69] Lankard, D.R. (1970). The dimensional instability of heated Portland cement concrete, Ph.D Dessertation, Ohio State University, Cleveland, OH. [70] Lee, J.S., Xi, Y., and Willam, K. (2006). Concrete under high temperature heating and cooling, submitted for publication to ACI Materials Journal.

229 [71] Leithner, D. (2004). Experimental investigation of concrete subjected to fire loading, Diploma Thesis, Technical University Vienna, Austria. [72] Lemaitre, J. (1992). A course on damage mechanics, Springer-Verlag, Berlin Heidelberg New York. [73] Lie, T.T., and Kodur, V.K.R. (1995a). Mechanical properties of fiberreinforced concrete at elevated temperatures, Institute for Research in Construction, National Research Council of Canada, Ottawa, Ontario, Internal Report No. 687. [74] Lie, T.T., and Kodur, V.K.R. (1995b). Thermal properties of fiber-reinforced concrete at elevated temperatures, Institute for Research in Construction, National Research Council of Canada, Ottawa, Ontario, Internal Report No. 683. [75] Lie, T.T., and Kodur, V.K.R. (1995c). Effect of temperature on thermal and mechanical properties of steel-fiber-reinforced concrete, Institute for Research in Construction, National Research Council of Canada, Ottawa, Ontario, Internal Report No. 695. [76] Lie, T.T., and Kodur, V.K.R. (1996). Thermal and mechanical properties of steel-fiber reinforced concrete at elevated temperature, Canadian J. Civil Engineers., 23, 511-517. [77] Lin, W.M., Lina, T.D., and Powers-Couche, L.J. (1996). Microstructures of Fire-damaged concrete, ACI materials journal, 93(3), 199-205. [78] Lu, N. and Likos, W.J. (2004). Unsaturated soil mechanics, Jone Wiley and Sons, Inc., New Jersey, USA.

230 [79] Ludirdja, D. Berger, R.L., and Young, F. (1989). Simple method for measuring water permeability of concrete. ACI material journal, 433-439. [80] Luo, X., Sun, W., Chen, S.Y.N. (2000). Effect of heating and cooling regimes on residual strength and microstructure of normal strength and highperformance concrete, Cement and Concrete Research, 30, 379-383. [81] Mehta, P.K. (1986). Concrete-Structure Properties and Materials, Prentice Hall. [82] Monteiro, P.J.M. and Chang, C.T. (1995). The elastic moduli of calcium hydroxide, Cement and Concrete Research, 25 (8), 1605-1609. [83] Morita, T., Saito, H., and Kumagai, H. (1992). Residual mechanical properties of high strength concrete members exposed to high temperaturepart 1. Test on material properties, Summaries of Technical Papers of Annual Meeting, Architectural Institute of Japan, Niigata, Aug. [84] Morteza, E.G., Xi, Y., Sture, S. (2005). A damage theory based on composite mechanics, Preparing for submitting to Journal of engineering mechanics. [85] Nakhi, A.E. (2004). Damage impact on chloride diffusion through concrete: experimental, theoretical, and numerical studies, Ph.D Dessertation, University of Colorado, Boulder, CO. [86] Nielsen, C.V., Pearce, C.J., and Bicanic, N. (2004). Improved

phenomenological modeling of transient thermal strains for concrete at high temperature, Computers and Concrete, March, 1(2), 189-209. [87] Noumowe, A.N., Clastres, P., Debicki, G., and Costaz, J.L. (1996). Thermal stresses and water vapor pressure of high performance concrete at high

231 temperature, Proceedings, 4th International symposium on utilization of High-strength/High-Performance Concrete, Paris, France. [88] Oda, H., Anderson, O.L., Isaak, D.G., and Suzuki, I. (1992). Measurement of elastic properties of single-crystal CaO up to 1200K, Physics and Chemistry of Materials, 19, 96-105. [89] Okajima, S. (1978). Study of thermal properties of rock-forming minerals, Master thesis of Okayama University. [90] Ozbolt, J., Kozar, I., Eligehausen, R., and Periskic, G. (2004). Transient thermal 3D FE Analysis of heated stud anchors exposed to fire, Workshop fib Task Group 4.3.2, Fire Design of Concrete Structures, Politecnico di Milano, Italy, Dec.2-4. [91] Pearce, C.J., and Bicanic, N. (2003). A transient thermal creep model for concrete, Proceedings of the EURO-C Conference, March 17-20, 2003, St Johann i.P., Austria. [92] Phan, L.T. (1996). Fire Performance of High-Strength Concrete: A report of the state-of-the art, Building and Fire Research Laboratory, National Institute of Standards and Technology, NISTIR 5934, Dec. [93] Phan, L.T., and Carino, N.J. (1998). Review of mechanical properties of HSC at Elevated Temperature, Journal of Materials in Civil Engineering, ASCE, Feb., 10(1), 58-64. [94] Phan, L.T., Lawson, J.R., and Davis, F.L. (2001). Effect of elevated temperature exposure on heating characteristics, spalling, and residual properties of high performance concrete, Materials and Structures (RILEM), March, 34, 83-91.

232 [95] Phan, L.T. (2002). High-strength concrete at high temperature-an overview, Utilization of High Strength/High Performance Concrete, 6th International Symposium, Volume 1, Leipzig, Germany, June, 501-518. [96] Phan, L.T. (2004). Codes and Standards for Fire Safety Design of Concrete Structures in the U.S., Workshop fib Task Group 4.3.2, Fire Design of Concrete Structures, Politecnico di Milano, Italy, Dec.2-4. [97] Philleo, R. (1958), Some Physical Properties of Concrete at High Temperatures, Journal of the American Concrete Institute, 29/54(10), 857864. [98] Piasta, J. (1984a). Heat deformations of cement paste phases and the microstructure of cement paste, Materiaux et Constructions, 17(102), 415420. [99] Piasta, J., Sawicz, Z., Rudzinski, L. (1984b). Changes in the structure of hardened cement paste due to high temperature. Materiaux et Constructions, 17(100), 291-296. [100] Pimienta, P. and Hager, I. (2002). Mechanical behaviour of HPC at high temperature, 6TH International Symposium on Utilisation of High Strength/High Performance Concrete, June, Leipzig, 16-20. [101] Poon, C.S., Azhar, S., Anson, M., and Wong, Y.L. (2001). Comparison of the strength and durability performance of normal- and high-strength pozzolanic concretes at elevated temperatures, Cement and Concrete Research, 31, 1219-1300. [102] Power, T.C. and Brownyard, T.L. (1947). Studies of the physical properties of hardened Portland cement paste, Journal of ACI, April, 984-988.

233

[103] Powers, T.C., and Brownyard, T.L. (1948). "Studies of the physical properties of hardened cement paste", Res. Lab. Portland Cem. Assoc. Bull. 22. [104] Powers, T.C. (1960). "Physical Properties of Hardened Cement Paste", In proceeding of the fourth international symposium on the Chemistry of Cement, 22, Washington DC, 577-609. [105] prEN 1992-1-1, Eurocode 2 (2002). Design of concrete structures Part 1: General rules and rules for buildings, Nov. [106] Ravindrarajah, R.S., Lopez, R. and Reslan, H. (2002). Effect of elevated temperature on the properties of high-strength concrete containing cement supplementary materials, 9TH International Conference on Durability of Building Materials and Components, Australia, 17-20th , March. [107] Reis, M.L.B.C., Neves, I.C., Tadeu, A.J.B., and Rodrigues, J.P.C. (2001). High-temperature compressive strength of steel fiber high-strength concrete, Journal of Material in Civil Engrg., ASCE, May/June, 13(3), 230-234. [108] Rigberth, J. (2000). Simplified Design of Fire Exposed Concrete Beams and Columns: An Evaluation of Eurocodes and Swedish Building Code, Report 5063, Department of Fire Safety Engineering, Lund University, Sweden. [109] RILEM Committee 68-MMH, Task Group 3. (1986). The Hydration of Tricalcium Aluminate and Tetracalcium Aluminoferrite in The Presence of Calcium Sulfate, Material and Structures, 19(10), 137-147. [110] Road Research Laboratory. (1959). Road stone test data presented in tabular form, DSIR Road Note No.24 (London, HMSO).

234 [111] Sanjayan, G., and Stocks, L. (1991). Spalling of High-Strength Silica Fume Concrete in Fire, ACI Spring Convention, Boston, MA, March. [112] Schneider, U. (1988). Concrete at high temperatures-A general review, Fire Safety Journal, 13, 55-68. [113] Schneider, U. and Herbst, H. (1989). Permeability and porosity of concrete at high temperature, Technical report 403, Deutscher Ausschuss fr Stahlbeton, Berlin, In German. [114] Schneider, U., and Herbst, H. (2002). Theoretical considerations about spalling in tunnels at high temperatures, Technical Report, Technical University Vienna, Austria. [115] Sidney, M., J. Francis, Y., and David, D. (2002). Concrete-Civil Engineering and Engineering Mechanics Series, Prentice Hall, 2nd Edition. [116] Soles, J.A and Geller, L.B. (1963). Experimental Studies Relating Mineralogical and Petrographic Features to The Thermal Piercing Rocks, Mines Branch Technical Bullitine, TB 53, Department of Mines and Technical Surveys, Ottawa. [117] Taylor, H.F.W. (1987). A method for predicting alkali ion concentration in cement pore solutions, Adv. Cem Res. 1(1), 5. [118] Taylor, H.F.W. (1997). Cement Chemistry, Academic Press, New York. [119] Tennis, P.D and Jennings, H.M. (2000). A model for two types of calcium silicate hydrate in the microstructure of Portland cement pastes, Cement and Concrete Research, 30, 855-863.

235 [120] Thelandersson, S. (1987). Modelling of combined thermal and mechanical action in concrete, J. Eng. Mech., ASCE, 113(6), 893-906. [121] Tsivilis, S., Kakali, G., Chaniotakis, E., and Souvaridou, A. (1998). A study on the hydration of Portland limestone cement by means of TG, Journal of thermal analysis, 52, 863-870. [122] Udeme J.N., and Bergeson, K.L. (1995). Thermal expansion of concretes: Case study in Iowa, Journal of Materials in Civil Engrg, ASCE, Nov.,7(4), 246-251. [123] Velez, K., Maximilien, S., Damidot, D., Fantozzi, G., and Sorrentino, F. (2001). Determination by nanoindentation of elastic modulus and hardness of pure constituents of Portland cement clinker, Cement and Concrete Research, 31 (4), 555-561. [124] Vishay Micro-Measurements. (2004). Measurement of thermal expansion coefficient using strain gage, Doc No. 11063, Technical note TN-513-1, VISHAY. [125] Wang, X.S., Wu, B.S., and Wang, Q.Y. (2004). Online SEM investigation of microcrack characteristics of concretes at various temperatures, Cement and Concrete Research, 35, 1385-1390. [126] Willam, K., Rhee, I. and Xi, Y. (2003). Thermal Degradation in Heterogeneous Concrete Materials, Proceedings of the EURO-C Conference, March 17-20, St. Johann i. P., Austria. [127] Willam, K., Basche, H.D. and Xi, Y. (2004). Constitutive Aspects of High Temperature Material Models, Workshop fib Task Group 4.3.2, Fire Design of Concrete Structures, Politecnico di Milano, Italy, Dec.2-4.

236 [128] Willam, K., Rhee, I., and Xi, Y. (2005). Thermal Degradation in Heterogeneous Concrete Materials, J. of Materials in Civil Engineering, ASCE, May/June, 17(3), 276-285. [129] Wittmann, F.H. (1986). Estimation of the modulus of elasticity of calcium hydroxide, Cement and Concrete Research, 16 (6), 971-972. [130] Xi, Y., and Jennings, H.M. (1997). Shrinkage of cement paste and concrete modeled by a multiscale effective homogeneous theory, Materials and Structures, 30 (July), 329-339. [131] Xi, Y. (2002). A Composite Theory for Diffusivity of Distressed Materials, Proc. of 15th ASCE Engineering Mechanics Conference, Columbia University, New York, June 2-6, No. 535. [132] Xi, Y. and Nakhi, A. (2005). Composite Damage Models for Diffusivity of Distressed Materials, Journal of Materials in Civil Engineering, ASCE, 17(3), 286-295. [133] Xi, Y., Eskandari-Ghadi, M., Suwito, and Sture. S. (2006). A damage theory based on composite mechanics, J. of Eng. Mech., ASCE, 132(11), 1-10. [134] Zeiml, M., Leitner, D., Lackner, R., and Mang, H.A. (2006). How do polypropylene fibers improve the spalling behavior of in-situ concrete? Cement and Concrete Research, 36, 929-942. [135] Zhukov V.V. (1975). Explosive failure of concrete during fire, Joint Fire Research Organization, Translation No. DT 2124, Borehamwood. England.

237

APPENDIX A

9.

RELATIVE HUMIDITY IN UNSATURATED SOIL

238
A.1. Definition of relative humidity

There are three assumptions in relative humidity in most practical geotechnical applications.
First assumption is that the composition of air excluding the water vapor

component remains essentially unchanged over time.

Second assumption is that the mixture of the component gases as a whole, as

well as each of the component gases that make up air, follows ideal gas behavior.
Third assumption is that all components of air, including the water vapor

component, reach local thermodynamic equilibrium. Thermodynamic equilibrium requires that chemical potentials among all components of all phase in the system are the same. R.H in the atmosphere is expressed as Eq. (A.1). From the ideal-gas equation of state,

T R P = R g = (T ) w

(A.1)

Where P is the absolute pressure, Rg = R w (gas constant), R is the universal gas constant, w is the molar mass (molecular mass), T is the temperature, = 1 is the specific volume, is the density. Notice: Definition of mol is that the mole (abbreviation, mol) is the Standard International (SI) unit of material quantity. One mole is the number of atoms in 12 g

239 of C-12 (carbon-12). This number is equal to approximately 6.022169 1023 , and is also called the Avogadro constant.

M Pw = V RT

(A.2)

Thus, the density can be expressed by Eq. (A.2). Relative humidity is defined as the ratio of the absolute humidity ( v ) in equilibrium with any solution to the absolute humidity in equilibrium with free water ( v ,sat ) at the same temperature.

P RH = v = RT = v v , sat Pv , sat wv Pv , sat

RT

Pv wv

(A.3)

Where v is the vapor density, v ,sat is the saturated vapor density, Pv is the partial water vapor pressure, Pv , sat is the saturated vapor pressure, wv is the molar mass (molecular mass) of water vapor, R is the universal gas constant, and T is the temperature in Kelvin.

A.2. Capillary tube model for unsaturated soil

A simple capillary tube model has been developed to analyses of unsaturated soil. The complex geometries for various shapes and sizes of particles and pore

240 fabrics formed among adjacent particles are simplified with the following assumptions. First assumption is that sand particles have identical spherical shape, second assumption is that an air-water interface described by the so-called toroidal approximation. The idealized geometry of the air-water interface between two spherical soil grains, which can be characterized by two radii of curvature r1 and r2, is shown in Fig. A.1.
A B Ts Soild A Water r2 A Soild Ts Ts A (a) Water meniscus between two particles (b) Free-body diagram for water meniscus T = Ts sin T d r3 r3 Air r1 r2 r1 r3 Ts B

(c) Section B-B

(d) Small area of section B-B

Figure A. 1. Idealized air-water interface geometry in unsaturated soil

Consider force balance in the horizontal direction and the free body diagram in Fig. A.1. There are three force contributions in the free-body diagram: surface

241 tension along the interface described by r1 that results in the positive direction horizontally, surface tension along the interface described by r2 that results in the negative direction horizontally, and air and water pressure applied on either side of the interface. The projection of surface tension in the positive horizontal direction is

F1 = 4 2 T cos r3d = 4 r3 T = 4 r3 Ts sin

0

(A.4)

The projection of the surface tension in the negative horizontal direction (if

is very small, = sin ) is

F2 = 2 Ts (2 r1 ) = 4 r1 Ts 4 r1 Ts sin

(A.5)

Also, the projection of air and water pressure Pa and Pw in the horizontal direction (assuming r2 = r3 ) is

F3 = ( Pa Pw ) (2 r1 sin ) (2 r2 ) = 4 r1 r2 ( Pa Pw ) sin
Balancing all three forces leads to
1 1 ( Pa Pw ) = Ts r1 r2

(A.6)

(A.7)

Eq. (A.7) provides a simple mathematical expression describing the pressure change across an air-water-solid interface between two idealized soil grains. The

242 quantity ( Pa Pw ) is called as the matric suction. The matric suction depending on the relative magnitudes of r1 and r2 can be positive, zero, or negative. Most likely, the value of matric suction is positive due to the fact that r1 is mostly less than r2 under unsaturated conditions. The matric suction can be also expressed in terms of soil pore radius ( r ), contact angle between air and water ( ), and surface tension Ts . Fig. A.2 shows two dimensional free-body diagrams for pressure and surface tension across a spherical phase interface between air and water.

r Ts Air phase Pa Pa A Water phase Pw Pw Ts Ts A R Ts

Figure A. 2. Free-body diagrams for pressure and surface tension across a spherical phase interface between air and water.

The projection of incremental force due to pressure on both sides of interface over an area A in the vertical direction is as follows:

Fv = ( Pa Pw ) A cos = ( Pa Pw ) A '

(A.8)

243 Where A' is the projection of A in the horizontal axis. The total vertical force due to the pressure difference acts over the area of the interface as follows:

Fv = ( Pa Pw ) r 2

(A.9)

The projection of surface tension around the circumference of the cut in the vertical direction is

Fv = 2 r Ts cos
Applying force equilibrium leads to ( Pa Pw ) = 2Ts cos r

(A.10)

(A.11)

A.3. Kelvins equation

Consider a simple three phase system comprised of air, water, and soil at a state of equilibrium in a closed container. The air phase consists of two components of dry air ( Pda ) and water vapor ( Pv ). The total air pressure ( Pa ) is equal to the sum of the partial pressures of dry air and water vapor. The composition and amount of dry air will not vary in the container, but the amount of water vapor may indeed vary under concurrent condensation and evaporation processes. Assume that the water phase is free (i.e. free of influence by the solid, the solid container, and dissolved solute) and that the air-water interface is perfectly flat (Fig. A.3).

244

Air Pa=Pda+Pv Water Solid

Figure A. 3. A simple three phase system comprised of air, water, and soil

Pw

For relatively incompressible materials such as solid, mechanical force considerations are usually the only criteria necessary to arrive at an equilibrium relationship. However, for highly deformable materials such as dry air, water vapor, or liquids, it is necessary to also consider chemical equilibrium. For this thought experiment, mechanical and chemical equilibrium between the air and water phases are considered. Because the air-water interface is flat, mechanical equilibrium requires that the air pressure ( Pa ) be equal to the total water pressure ( Pw ). Chemical equilibrium requires that the total chemical potential, or more conveniently, the change in the total chemical potential, be the same in each coexisting phase (i.e., air and water).

For mechanical equilibrium,

Pa = Pda + Pv = Pw
For chemical equilibrium,

(A.12)

a = da + v = w = RT = Pwvw = Pa va = Pda vda + Pv vv

(A.13)

245

Where i and vi mean the chemical potential and the partial molar volume for each component, respectively. Assuming ideal gas behavior for the dry air and water vapor, the dry air pressure and vapor pressure can be expressed as Eq. (A.14) and (A.15) respectively.

Pda =

M da RT RT = Vda wda vda Mv RT RT = Vv wv vv

(A.14)

Pv =

(A.15)

Where M is the mass, V is the volume, and w is the molecular mass. Since any ideal gas has 22.4 L/mol, the partial molar volume can be calculated from molecular weight and the volume fraction of each respective gas.

From chemical equilibrium,

w = Pda vda + Pv vv

(A.16)

At mechanical and chemical equilibrium, the vapor pressure of pure water reaches its saturated value Pv , sat under the prevailing temperature and pressure condition. In other words, a state of 100 % relative humidity is reached. This state is defined as reference state.

246

Air

Pa=Pda+Pv

Water Droplets, Pw

r
Solid

Figure A. 4. Air-water-solid system at mechanical and chemical equilibrium

Lets suppose that all of the water in the container is in the form of spherical droplets having uniform radii r (Fig. A.4). The solid container consists of a perfectly water repellent material such that the contact angle 180 , implying that no water potential change can occur due to surface wetting. Here, a new state of pressure and potential for the air and water phase must be established. If water vapor follows the ideal gas law and the change in chemical potential of dry air is negligible compared to the change in chemical potential of the water vapor, the change in chemical potential for the total air phase with respect to the previous case for the flat air-water interface can be expressed by Eq. (A.17).

a = da + v = v = RT ln

Pv Pv , sat

(A.17)

The last assumption in the development of above the equation is based on the fact that the total pressure change is small and that the partial molar volume of dry air remains unchanged in the closed container.

247 The chemical potential change in the liquid phase is expressed by Eq. (A.18).

w = vw Pw

(A.18)

In the geometry of the water droplets (contact angle 0 between water and air: Fig. A.4.), the pressure change across the air-water is expressed by Eq. (A.19).

Pw = Pa Pw =
w = a = RT ln

2Ts r

Thus,

(A.19)

Pv 2T v = vw ( Pa Pw ) = s w Pv , sat r

or

(A.20)

P 2T RT ln v = s = ( Pa Pw ) vw Pv , sat r

(A.21)

Eq. (A.21) is Kelvins equation applied to equilibrium between a water drop and its vapor pressure.

Air

Pa=Pda+Pv

Capillary Tubes, Pw r Solid

Figure A. 5. Air-water-solid system at mechanical and chemical equilibrium

248

To apply to unsaturated soil, consider the idealized system of capillary tubes partially filled with water, each having a radius r and a solid-liquid contact angle (Fig. A.5). From the geometry of Fig. A.5, Kelvins equation can be expressed as Eq. (A.22).

w = a = RT ln

Pv 2T v cos = s w Pv , sat r

(A.22)

The relative humidity of the pore air phase in saturated soil is fundamentally linked to the change of chemical potential and the pressure difference between air and water (suction). Eq. (A.23) is obtained from Eq. (A.11) and (A.22) using the capillary tube model. Finally, the relative humidity in unsaturated soil is expressed with Eq. (A.24).

Pa Pw =

P 2T cos RT RT ln v = ln ( RH ) = s vw Pv ,sat vw r

(A.23) (A.24)

( P - P )v 2T cos vw RH = exp - a w w = exp - s RT r RT