You are on page 1of 23

Performance Based Fire Safety Design of Structures A Multi-dimensional Integration

J. Y. Richard Liew*
Department of Civil Engineering, National University of Singapore, Blk E1A, #05-13, 1 Engineering Drive 2, Singapore 117576 (Received 16 July 2003; Received revised form: 17 November 2003; Accepted: 18 November 2003)

Abstract: Design codes for fire safety in buildings can be either a prescriptive type or performance-based type. It is now widely recognized that performance-based codes provide greater advantages over the prescriptive codes in that it allows designers to use the fire engineering methods to assess the fire safety of the structure. However, as the assessment of the whole structure performance is not easy, most codes currently used are still prescriptive codes or a combination of prescriptive codes and performancebased codes. The key feature for implementing the performance-based fire design codes is the assessment of the fire resistance of the structure. This paper provides an overall view on performance-based code and the approaches for designing steel structure in fire considering a multi-dimensional integration of fire engineering simulation, emergency evacuation and structural resistance. Various fire models and heat transfer analysis methods are reviewed and discussed. The basis to modelling of large deflection and plasticity using appropriate stress-strain relationship at elevated temperature is explained. Finally, structural response calculations from simplified hand calculation method to advanced numerical procedures are presented. Future trends for research are identified.

Key words: fire safety, emergency evacuation, explosion, fire modelling, fire protection, performance-based design, tall buildings,
plastic hinge analysis.

1. PERFORMANCE BASED DESIGN FOR FIRE SAFETY


A rational approach to fire safety assessment is to relate functional requirements, such as prevention of spreading heat and smoke, safe evacuation and rescue etc., to fire resistance considering both local and global stability of structures. In a performance-based design, the designer needs to first understand the level of performance is expected, then to design to these levels and finally to predict the performance that will be achieved to ensure the reliability and robustness of the design. At the beginning of the performance-based analysis and design, the designer and the stakeholders should both agree on the project scope, intent and building use.

The designer will then define and prioritize the fire safety plans and performance criteria and define the fire safety goals and objectives pertaining to the design intent. The next step to follow is to determine the implementation of the selected solutions into the project scope and execution. The final stage is to establish the necessary verification and quality assurance to ensure compliance with the agreed solutions. In general, a process for performance-based design would involve the following steps: 1. identify goals and defining stakeholder and design objectives 2. identify the possible fire scenarios which could occur during the service life of the structure,

*Corresponding author. Email address: cveljy@nus.edu.sg; Fax: +65-6779-1635; Tel: +65-6874-2154.

Advances in Structural Engineering Vol. 7 No. 4 2004

311

Performance Based Fire Safety Design of Structures A Multi-dimensional Integration

evaluate the likelihood and consequences of such scenarios, 4. establish appropriate performance criteria by ensuring effective evacuation, escape and rescue and to prevent injury arising from such events, 5. predict the performance of the system based on available engineering data, and 6. ensure robustness of the designs, reliability and durability of the protection systems, etc. Although the governing criteria for impairment of the main safety functions are often non-structural (for example exposure to heat, temperature, toxic gases etc.), it is, however, a functional requirement that the structure would remain stable to allow adequate time for safe evacuation and rescue. Therefore a quantitative assessment on fire resistance of a structure is necessary and this opens up the opportunity in using advanced simulation tools and computational methods developed as reported in the recent fire workshops (Liew 2002; Moss 2002).

3.

2. EMERGENCY EVACUATION
The objective of providing emergency evacuation is to allow occupants to travel safely to a place of safety in the event of fire. In general, the procedure involved in estimating evacuation time is given below: 1. define the space domain to be analyzed 2. estimate the response time of occupants 3. calculate the travel time to a specific location 4. determine whether this location is safe 5. if the specific location is safe, record the total required safe escape time and make sure that it does not exceed the available safe escape time, which is determined by untenability condition. If the specific escape route is not safe, select another one and repeat Step 3. A range of possible emergency scenarios may occur in high-rise and complex buildings for which it is necessary to develop a range of strategies for managing accidental fires. The following points are considered when planning for an evacuation strategy for such systems: Risk perception in high rise/complex buildings Types of fire scenarios that may occur and may need different evacuation strategies Fire scenarios that will affect building protection strategies and the emergency plans proposed Limitations of each emergency plans Integration of fire service into the emergency plan Types of information to be provided to the occupants Emergency evacuation training and fire protection maintenance plan

A major limitation on prediction of incident outcomes for performance-based design and hazard assessment is the lack of quantitative data on pre-movement time and evacuation time, and occupant behaviours for different fire scenarios and occupancies. There is a need to study human behaviour during evacuation from a range of occupancies. Advanced computer software is available to predict the evacuation time considering the effects of exposure to fire effluence on occupant evacuation behaviour. However, improvements are still needed to include the effects of different warning systems, information provided to occupants, occupant characteristics, pre-training, building complexity and level of fire safety management on pre-movement and evacuation time. The September 11 incident had shown that fire service intervention occurred while there were many occupants still in the buildings, many were in the process of evacuating while other remained in refuge were trapped by fire. In order to achieve effective design for buildings in emergency situations, it is essential to consider occupants characteristics, their abilities to evacuate and the effects of exposure to fire effluence on occupant evacuation behaviour.

3. FIRE MODELLING
Fire modelling is a mathematical simulation of the fire conditions in a compartment and is capable of giving information based on the parameters which have been designed. The fire development in a room normally involves three phases: pre-flashover, post-flashover and fire decay. In the pre-flashover phase, fuels begin to burn and the gas temperature varies from one point to another in the compartment. In the post-flashover phase, the fire develops fully, and the gas temperature increases rapidly to a peak value and becomes practically uniform throughout the compartment. The fire has the most influence on structural design because of high temperature and radiant heat fluxes produced in this phase. In the fire decay phase, the available fuel begins to decrease and the gas temperature falls. There is considerable benefit when the effects of natural fires in buildings, where the amounts of the combustible contents are small and the buildings are of large volume, is considered than using the standard ISO fire. For many years, fire engineering research has shown that overall structure performs better than isolated members in a fire situation. Numerous studies have been carried out to determine the temperature reached in real (natural) fires, to quantify the factors that govern fire severity and to investigate the parameters that cause structures to fail in fire. The studies show that the severity of natural fires in building compartments is governed by the amount of combustible material (the fire load), the

312

Advances in Structural Engineering Vol. 7 No. 4 2004

J. Y. Richard Liew

area of the doors and windows (the ventilation), and the thermal characteristics of the wall, floor and ceiling materials. In addition, fire-fighting measures are also important for the determination of fire exposure. The following subsections discuss the various models available for modelling fire of different complexity and severity. 3.1. Design Fire Models Design fires are derived empirically and may yield reasonable and consistent predictions provided that the fire conditions are similar to those in the underlying assumptions. Standard fire curves such as ISO-834 do not represent the real fire in a compartment and serve only as criteria to evaluate the fire resistance capacity of single structural members. Simplified code prescribed method often assumes that the fire has a constant temperature throughout the burning period. Eurocode 1: Part 1-2 (2001) recommends equations for parametric fires, allowing a temperature-time curve to be produced for any combination of fuel load, opening factor, height of opening and thermal characteristic of the boundary materials. The temperature T (C) and time relation during the heating phase is given as (Eurocode 1: Part 1-2, 2001): T = 20 + 1325 1 0.324e 0.2 t * 0.204e 1.7t * 0.472e 19 t *

qt,d is the design value of the fire load per total surface area At of the enclosure. The recommended fire growth rate is taken as tlim = 25 minutes, 20 minutes and 15 minutes for slow, medium and fast growth rate, respectively. The introduction of tlim is to avoid unrealistic short fire duration when the ratio between the fire load and the opening factor decreases. Any object or fire load needs a certain amount of time to burn, even if there is unlimited presence of air (Franssen 1997). The temperature-time curve during the cooling phase is given by: T = Tmax 625(t* t*max.x) for t*max 0.5

T = Tmax 250(3 t*max)(t* t*max.x) for 0.5 < t*max < 2 T = Tmax 250(t* t*max.x) in which t* = t. t*max = (0.2 103 qt,d /O). x = 1.0 x = tlim./ t*max if if tmax > tlim tmax = tlim for t*max 2 (6)

(1) where t* is the fictitious time given by t* = t. t is the time (hr) and = (2)

(0.04 /1160)2

(O / b ) 2

(3)

where b is the square root of thermal inertial of the boundary material of the compartment and O is the opening factor (m1/2) given by O = Av heq At (4)

Av is the total area of vertical openings on all walls; heq is the weighted average of window heights on all walls and At is the total area of enclosure (walls, ceiling and floor, including openings). In case of = 1, Eqn 1 approximates the ISO834 standard temperature-time curve. Depending on whether the fire is fuel controlled or ventilation controlled, the duration of the heating phase tmax (hr) is given as: t max = Maximum 0.2 10 3 q t,d O , t lim

Figure 1 shows the parametric fire curves plotted for a range of opening factors (OF), fuel loads and materials according to the Eurocode 1: Part 1-2 (2001). Fire curves are produced for three fire loads, four opening factors and two types of construction, showing the significant dependence of fire temperature on the bounding materials. The fire loads are 400, 800 and 1200 MJ per floor area, for a room 5 5 m in plan and 3 m high. Feasey & Buchanan (2002) pointed out that the Eurocode equation gives extremely fast decay rates for large openings in well insulated compartments (e.g., OF = 0.12 in Figure 1) and extremely slow decay rates for small openings in poorly insulated compartments. They proposed some modifications to the Eurocode formula to give a better estimation of the temperature-time curve in the fire decay phase. 3.2. Zone Models Zone models represent more of the phenomenological behaviour of fire. They solve the conservation equations for distinct and relatively large regions. In each zone, the heat balance equations are solved to generate gas temperatures. There are several options for calculating the heat release rate, based on ventilation control, fuel control or the porosity of wood crib fuels. Other computer models including ZONE, CSTBZ1, CFAST, BANZFIRE,

(5)

Advances in Structural Engineering Vol. 7 No. 4 2004

313

Performance Based Fire Safety Design of Structures A Multi-dimensional Integration

1400

Temperature (C)

1000 800 600 400 200 0 0 1400

ISO 834 Concrete

Temperature (C)

1200

Gypsum

OF = 0.02

1400 1200 1000 800 600 400 200 0 0 1400

Gypsum

OF = 0.04 ISO 834

Concrete

50 100 150 200 250 300 350 400

50 100 150 200 250 300 350 400

Time (min)
1200 1000 800 600 400 200 0 0 20 40 60 80 100 120

Time (min)
1200 1000 800 600 400 200 0 0 20 40 60

Temperature (C)

Temperature (C)

OF = 0.08 Gypsum ISO 834 Concrete

Gypsum

OF = 0.12 ISO 834

Concrete
80 100 120

Time (min)

Time (min)

Figure 1. Parametric temperature-time curves (fuel load = 400, 800, 1200 MJ/m2 floor area)

Figure 2. Multi-zone fire model

are summarised in SFPE (2002). Schleich (1996) developed a realistic fire evolution model, which not only takes into account the physical factors, but also the influence of active protection measures in the structure. The multi-zone fire model (Liew et al. 2002) places heating boxes one outside another and assumes uniform heating in each box, as shown in Figure 2. The smallest

box nearest to the heat source has the highest temperature. The boxes further from the heat source have lower temperature, and the farthest box has the lowest temperature. The structural elements enclosed within each heating box are subjected to a uniform heating rate, which can be either constant or vary as a function of time. The temperature-time relationship in each box

314

Advances in Structural Engineering Vol. 7 No. 4 2004

J. Y. Richard Liew

may be obtained by calibration with fire tests or from numerical simulation based on the theory of thermal energy balance. The air temperature at each time step can be prescribed. The multi-zone fire model (Liew et al. 2002) provides a means to calibrate actual fires. The heat intensity and the flame size of the actual fires depend on many factors such as fire load, ventilation, active fire control devices, and so on. If there is such a part in a structure that is heated significantly more or less than others, the multi-zone model could be a more suitable fire model to simulate an open fire. The multi-zone fire model can be used to simulate uniform heating. In this case, only one-box is prescribed and all structural members within this box will be heated simultaneously under the same fire load. 3.3. Radiation Model Radiation model may be used to simulate fire as a radiating source with the heat flux intensity defined by its distance from the source, as shown in Figure 3. The heat rays are emitted in all directions. The heat intensity may be constant or vary as a function of time. The heat flux intensity, q, received by the individual elements is calculated as (SINTEF 1995): qi = E cos i 4ri2 (7)

known (Yao et al. 1995):

q Tk = i

14

(8)

where = the emissivity coefficient; qi = the heat flux intensity calculated from Eqn 7; = the Stefan-Boltzman constant of 5.67 108 W/m2K4; TK = the temperature on the spherical surface with a unit of Kelvin (K). The radiation model is a convenient way to propose a simple relationship between the air temperature and its distance from the fire source. For large and complex structures, the radiation model is preferred as it is relatively simpler to use for simulating an open fire. On the other hand, the multi-zone fire model requires a descritization of space into finite boxes. Each heating box needs to be prescribed with an appropriate timetemperature relationship. 3.4. Computational Fluid Dynamic (CFD) Approach The CFD model can be used to represent various types of walls of different materials and the exact location and dimensions of openings. The fire compartment to be modelled by CFD is divided into a number of small volumes. The fluid dynamics equations are written in each of these small volumes, and each volume is linked to the adjacent volumes. The heat transient problems are expressed in differential equations, and time integration has to be performed by solving a large number of equations in the time domain. The main drawback of CFD model is that a significant number of parameters have to be given and many of them are variable with very little, if any, link to any physical phenomenon. It requires an experienced user before the any meaningful
Intensity

where Ei = total energy emitted from the source ri = distance between the heat source and the midpoint of ith element i = angle between the ray and the element surface normal (see Figure 3). The air temperature can be computed by adopting the Stefan-Boltzman formula once the heat flux intensity is

rad ius
(x, y, z)

z y x I = intensity cos() Figure 3. Radiation model to simulate open fires

Advances in Structural Engineering Vol. 7 No. 4 2004

315

Performance Based Fire Safety Design of Structures A Multi-dimensional Integration

Plot 3d Speed m/s

8.00 7.20 6.40 5.60 4.80 4.00 3.20 2.40 1.60 0.80 0.00

Figure 4. Example of smoke velocity in a fire compartment from CFD model

result can be obtained. Practising engineers often find difficulties in understanding the hypothesis underlying the CFD models and they have to be careful in interpreting the results. A large amount of results can be produced by CFD computation. For instance, it provides the temperature components of the velocity, the pressure, the oxygen concentration, etc, in every volume at every time step. A graphical representation of the results is mandatory, see Figure 4. To some extend, the amount of results can even create a difficulty if the interest is for the behaviour of structure. An interface is often needed to be created between the CFD model and the structural model, because thermal environment in the compartment calculated by the CFD model would influence the structural element. The main advantage of CFD models is to model compartments with complex geometry because only the flexibility given by the fine discretisation of the space inherent to CFD models allows a correct representation of this complexity. Notwithstanding the costs and problems associated to these models, they are favoured by researchers who use them as the design tools for fire analysis. Their application domain is for complex projects which are very often combined the geometrical complexity and the financial resources which make the CFD model an attractive choice.

the member. Calculation of member temperature requires solution in time domain via a fairly complex differential equation. There are two kinds of heat transfer methods used in fire engineering design: analytical method and finite element method. 4.1. Analytical Method Different standards or specifications give simplified way to calculate the net heat flux and temperature development in the steel member. The heat flux due to convection is proportional to the temperature gradient between the ambient gas temperature and temperature of the steel member. The heat flux due to radiation is proportional to the temperature gradient of the forth order of the ambient gas temperature and the steel temperature. ECCS (1993) uses a single expression to represent the total heat flux as = h A ( ) Q s s f (9)

4. HEAT TRANSFER ANALYSIS


The process of heat transfer between a fire and a structure can be described by the balance between the net incident thermal radiation and convective heat flux and the rate of heat conducted in the material. The rate of heating of any structural member is dependent at any time on the temperatures of both the fire atmosphere and

where As = surface area of the member per unit length exposed to heating f = fire temperature at time t s = temperature of the steel member h = coefficient of total heat transfer Based on this heat flow law, the temperature development in unprotected steel member can be calculated. Eurocode 3: Part 1-2 (2001) provides a rational means to estimate steel temperature development by considering the section factor and configuration factor for internal steelwork and external steelwork. The temperature increments in the structural members are calculated over small time steps. This method is particularly suitable for calculation using simple spreadsheet programming.

316

Advances in Structural Engineering Vol. 7 No. 4 2004

J. Y. Richard Liew

A = area of the steel section B

t D D

Am/V = 2(2B+Dt)/A

Am/V = (3B+2D2t)/A

Am/V = 2(B+D)/A

Figure 5. Values of parameter for use in the calculation of Section factor Am/V

For internal steelwork, the increase of temperature (s) in an unprotected steel section during in a time interval t may be calculated as (Eurocode 3: Part 1-2 2001): s = kshadow Am V hnet ( f s )t cs s (10)

section factors for selected examples are shown in Figure 5. For protected steel members under fire, the protection material of low thermal conductivity reduces the rate of heat transfer from the fire to the steel section. The increase in steel temperature s in a time increment t due to the heat transfer from the fire through the fire protection to the steel section may be calculated as: s = A p / V( f s ) ( t p / k p )css (1 + /3) t ( e /10 1) f (13)

in which kshadow = correction factor for the shadow effect of flanges, defined as a ratio of the box value of the section factor to the section factor cs = specific heat of steel [J/kgK] s = density of steel [kg/m3] Am / V = section factor for unprotected steel members hnet = Heat transfer coefficient per unit area per degree Kelvin t = The time interval [seconds] On the fire exposed surface the heat transfer coefficient hnet may be determined by considering heat transfer by radiation and convection as hnet = hnet ,c + hnet ,r where the radiative term may be expressed as: hnet ,r = r ( f + 273)2 ( s + 273)2 f + s + 546 (11)

in which the relative heat storage in the protection material is given as = c pp css tp Ap V (14)

]
(12)

and A p /V = section factor for protected steel member, where A p is the inner perimeter of the protection material. cs , c p = specific heats of steel and protection material t p = thickness of fire protection material s , f = temperatures of steel and fire at time t f = increase of fire temperature during the time step t k p = thermal conductivity of the fire protection material s , p = densities of steel and fire protection material The temperature development in protected steel members can be evaluated based on the thermal properties of the insulation materials. 4.2. Finite Element Method Finite element method may be used to estimate the thermal effects on the structural elements by subdividing the structural element into a number of quadrilateral heat transfer elements. Heat conduction, heat convection and exchanges of radiation are calculated on the basis of the heat transfer element. One simplified approach is to store the temperature history in each structural member and then calculate the equivalent nodal expansion based on the incremental temperature change. Consistent nodal forces are produced on an elastic element at elevated

where = Stefan-Boltzmann constant of 5.67 10 8 r = f m = resultant emissivity using f = 1.0 and m = 0.7, where f and m are the emissivity of the fire and the surface of the member, f and s = Fire and steel temperature, respectively, in Celsius. The convective heat transfer coefficient is hnet,c = 25 W/m2K for standard fire, 35 W/m2K for natural fire and 50 W/m2K for hydrocarbon fire. The section factor Am / V uses the fire exposed perimeter in calculating an appropriate value of Am . The

Advances in Structural Engineering Vol. 7 No. 4 2004

317

Performance Based Fire Safety Design of Structures A Multi-dimensional Integration

temperatures due to the axial expansion and temperature gradient increment over the cross-section. These forces can be used in the structural analysis to determine the responses of the structure in fire (Ma and Liew 2004). Although the techniques for solution of the heat transfer problem are relatively well established, several complicating factors exist. For examples, physical properties such as thermal conductivity, specific heat, emissivity/absorbtivity and heat transfer coefficients vary with temperature. Surfaces may also be exposed to re-radiation from other surfaces. The fraction of the emitted radiation received is governed by the configuration factor, which may be tedious to calculate, especially if shadowing of other members are present. Numerical method is often used to improve the accuracy of the heat transfer analysis.

stress. The temperature dependence of the proportional limit, the effective yield strength as well as the elastic modulus recommended by the code is shown in Figure 7. Creep may be of importance in a fire situation where a cross section is subjected to high temperature (above 400C) and high stress for a long period of time. The stress-strain curves given in the Eurocode code are based on measurements at constant temperature within a period of time to allow creep to take place in the tests. Therefore, creep effect is implicitly included in the effective yield strength used for design. In other words, when the design is based upon code values, creep does not need to be considered explicitly. If the temperature and load history is such that a structural component remains at high temperature or is highly stressed for only a short period, the prediction using the code values should yield conservative results.

5. MATERIAL BEHAVIOUR AT ELEVATED TEMPERATURE


Experimental evidence shows that the stiffness and strength of steel deteriorate at elevated temperatures. Typical stress-strain curve of steel at elevated temperatures is shown in Figure 6. The stress-strain relationship at elevated temperature does not exhibit a distinctive yield plateau. Therefore, the yield stress, or 0.2% proof stress, which is conventional design strength for steel at ambient temperature, loses its relevance because of the nonlinearity of the stress strain curve. Since fire is considered to be an accidental situation, large plastic strains are allowed. Hence, an effective yield stress is used, which is attainable when the strain is considerably larger than the elastic limit at normal temperatures. Eurocode 3: Part 1-2 (2001) adopts a yield strain y, of 2% to define the effective yield

6. NONLINEAR ANALYSIS AT ELEVATED TEMPERATURE


Modern design standards such as Eurocodes provide sufficient guidance to assess the fire performance of individual members in a fire compartment of a building framework. In the case of a braced frame in which each storey comprises a separate fire compartment with sufficient fire resistance, the effective buckling length of a column may be used to compute the limit load of the frame. However, guidance is not given for sway frames in which storey buckling and overall stability may dominate the design of individual member. Wang et al. (1995) provide simplified methods to analyse the performance of steel frames. They study the effect of continuity on the fire resistance of columns in both sway and non-sway steel frame and suggested some restraint f y, h effective yield strength;

Stress q

f p, h proportional limit; E a, h slope of the linear elastic range; x p, h strain at the proportional limit;

f y, h f p, h

x y, h yield strain; x t, h limiting strain for yield strength; E a, h = tan a


a

x u, h Ultimate strain;

x p, h

x y, h

x t, h

x u, h

Strain x

Figure 6. Stress-strain relationship of steel at elevated temperature according to European Committee for Standardisation (2001)

318

Advances in Structural Engineering Vol. 7 No. 4 2004

J. Y. Richard Liew

Reduction factor, k
1.0

Effective yield strength


k y,h = f y,h / f y

0.8

0.6

Design strength for satisfying deformation criteria


k x,h = f x,h / f y

0.4

Slope of linear elastic range


0.2

Proportional limit
k p,h = f p,h / f y

k E,h = E a,h / E a

0 0 200 400 600 800 1000 1200

Temperature [C]

Figure 7. Reduction factor for steel according to Eurocode 3: Part 1-2 (2001)

stiffness values at the end of the column due to the continuity of the sub-frame from the parametric study. General commercial programs such as ABAQUS, ANSYS, NASTRAN, may be used for analysing structures exposed to fires. They offer the advantages of full validation, powerful ability to model different kinds of problems and availability of further development so that they can almost suffice all needs. But these programs are rather inconvenient to use for being both time-consuming and complicated to operate since they are general purpose program not specifically written to perform analysis of structures under fire condition which is highly non-linear and transient. One powerful tool for analyzing large-scale steel structures exposed to fire is to adopt a second order plastic hinge-based analysis (Liew et al. 1998, 2002). In this approach, it is assumed that cross section is compact and the full plastic capacity can be achieved. At elevated temperature the plastic strength surface should follow the effective yield strength and its temperature reduction curve k y, for yield, as illustrated in Figure 7. The elastic modulus also degrades at elevated temperature following the temperature degradation curve for the slope of linear elastic range kE , as in Figure 7. Other fire effects include thermal expansion and thermal bowing. Further improvement to this model is to model the gradual plastification of members cross section using a twosurface plastic hinge method, which captures the gradual yielding of cross sections at elevated temperature (Ma and Liew 2004; Liew et al. 2000).

The two-surface plastic hinge model, which is formulated based on the bounding surface plasticity concept, represents the inelastic cross section behaviour by considering the interaction of axial force and bi-axial bending. The initial yield surface is assumed to be a scaled down version of the bounding surface that is fixed in size and translates without rotation in a stressresultant space. The gradual translation of the initial yield surface towards the bounding surface provides a smooth transition from initial yield to full plastification of cross section. Moreover, the element displacement fields are derived from the exact solution of the fourth order differential equation for a beam-column subjected to end forces (Liew et al. 2000), hence it is accurate enough to use only one beam-column element to model the stability behaviour of column member. At elevated temperature, the yield surface and the bounding surface have to contract in size in order to satisfy the yield condition. The degradation of the yield strength is based on the effective strength concept. At high temperature, the stress-strain relationship of steel is highly nonlinear and does not exhibit a distinct yield plateau. The idea of the effective strength is introduced to define a yield plateau at a relatively high strain level. For the two-surface plasticity model, the size of the bounding surface corresponding to full cross-sectional plastification, follows the reduction curve for effective yield strength in Eurocode 3: Part1-2 (2001), as illustrated in Figure 8. The size of the initial yield surface is assumed to degrade proportional to the bounding surface.

Advances in Structural Engineering Vol. 7 No. 4 2004

319

Performance Based Fire Safety Design of Structures A Multi-dimensional Integration

1 0.9 0.8 bounding surface (200C)

Strength degradation Ratio

0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 200 400

initial yield surface (200C)

effective yield strength bounding surface (700C) initial yield surface (700C)

initial yield strength

600

800

1000

1200

Temperature [C] Figure 8. Size of yield and bounding surface at elevated temperature according to Eurocode 3

q = 15 KN/m

254 146 UB

Uy 4.5 m Figure 9. Model 1: simply supported beam (4-side heated)

Ux

However, it can be observed from Figure 8 that the initial yield strength decreases at a faster rate than the effective yield strength; therefore, the size of the initial yield surface may be over predicted at higher temperature. In practice, this will not have any significant effect on the inelastic behaviour of members in fire. The verification of the two-surface plasticity model at ambient and elevated temperature is reported in Liew et al. (1998) and Ma and Liew (2004). Verification studies have been carried out on both components and frames over a wide range of parameters including uniformly heated members, three-side heated members with concrete slab acting as heat sink, members with passive fire protection and 2-D frames. Several examples are given in the next section to illustrate the application of the plasticity model and to study the accuracy of the model in modelling the inelastic behaviour of frame structures.

7. COMPARISON OF PLASTIC HINGE METHOD WITH SPREADOF-PLASTICITY ANALYSIS


Figures 9 to 11 show a simply supported beam, a single storey braced frame and a multi storey braced frame subjected to fire. The main purposed of the study is to compare the results obtained from fire analyses based on the plastic hinge (P-Hinge) method and the Spread of plasticity (S-Plastic) methods. All the structural models are subject to ISO standard fire. For the simply supported beam (model 1), the member cross section is exposed to 4-side fire (Figure 9). The temperature is assumed to be uniform over the cross section and along the length. The second and the third models are based on the portal frame shown in Figure 10. In model 2, the frame members are exposed to fire from all sides without fire protection while in model 3, the beam is protected by concrete slab,

320

Advances in Structural Engineering Vol. 7 No. 4 2004

J. Y. Richard Liew

P = 500 kN

q = 25.4 kN/m

P = 500 kN

Ux

Uy

Beam section: 305*165UB40 Column section: 203*203UB52 S275 steel

3.5 m

Columns are 4-side heated

5.5 m Figure 10. Rigid portal frame: Model 2: beam is 3-side heated; Model 3: beam is 4-side heated

P = 75.5 kN

P = 151 kN

P = 151 kN

P = 75.5 kN

3m

U.D.L = 25.4 kN/m over all beams

Beam section: 305*165UB40

3m

Ux Uy
Column section: 203*203UB52 S275 steel

3m

5.5 m

5.5 m

5.5 m

Figure 11. Model4: three-bay, three-storey rigid frame

and the beam section is 3-side heated. Model 4 is a 3-bay, 3-storey rigid frame (Figure 11), in which the lower left corner compartment is subject to fire. S275 steel with yield strength of 275 N/mm2 is used for all the structural members. For model 1, only vertical deflection of the beam is compared. For the other models, the comparison is made both on the mid-span deflection of the beam and the lateral deflection at the top of the heated column. The results are shown in Figures 12 to 15. In all figures, solid lines represent the vertical deflection of the beam, and the dashed line means the lateral deflection of the column. The comparison of the results indicates that the plastic hinge method gives satisfactory results before the formulation the first plastic hinge in the structure. If the failure is not due to the formation of plastic hinge

mechanism, then, the plastic hinge method shows a greater stiffness reduction than the spread-of-plasticity method (Figures 12 and 13). If a collapse mechanism occurs (Figures 14 and 15), the plastic hinge method shows collapse with rush out of deformation while the spread-of-plasticity analysis shows ability to sustain further load with moderate deformation. Hence the plastic hinge method may be used to predict the collapse of structures. But it may underestimate the post-collapse stiffness of the structures. For steel structures under fire attack, it is possible to allow the structure to undergo large deformation as long as the structure maintains stable. Therefore, when postcollapse behaviour is needed, the spread-of-plasticity method should be used.

Advances in Structural Engineering Vol. 7 No. 4 2004

321

Performance Based Fire Safety Design of Structures A Multi-dimensional Integration

700 600 500

Solid line: Uy Dash line: Ux

Deformation (mm)

400 300 200

P-Hinge
100 0 100

S-Plasticity

10

15

20

25

Time (minute) Figure 12. Deformation of structural model 1

1600 1400 1200 1000

uy (mm)

PHinge

800 600 400 200 0 0 2 4 6 8 10 12 14 16

Formulation of the plastic hinge at the mid of the beam

SPlasticity

Time (minute) Figure 13. Deformation of structural model 2

8. VERIFICATION OF NUMERICAL METHODS


Verification studies are important to ensure the validity of any analytical or numerical methods. Ma and Liew (2004) established the accuracy of their proposed advanced analysis method against experimental results for both individual members and complete frames exposed to fire. 8.1. Uniformly Heated Members Critical temperatures from the advanced analysis are compared with those in BS5950: Part 8 (BSI 1990) for

beams under uniform distributed loads and columns under axial loads in Table 1. Critical temperature for beams is taken at a mid-span deflection of L/20 and critical temperature for columns is taken at the failure of the column symbolized by a sudden increase of lateral deflection. The results agree well with each other. 8.2. Three-side Heated Beams Eighteen UK standard fire tests on unprotected simply supported steel beams supporting concrete slabs without

322

Advances in Structural Engineering Vol. 7 No. 4 2004

J. Y. Richard Liew

400 350 300

Solid line: Uy Dash line: Ux

Plastic hinge of the beam Deformation (mm)


250

P-Hinge
200 150

yielding of the column


100 50 0 50 0 5 10 15 20 25

S-Plasticity

Time (minute)

Figure 14. Deformation of structural model 3

300

Solid line: Uy Dash line: Ux


250

Deformation (mm)

200

150

P-Hinge
100

S-Plasticity
50

0 0 2 4 6 8 10

12

Time (minute) Figure 15. Deformation of structural model 4

composite action are simulated. The test descriptions and data are available in Wainman (1988). Table 2 summarizes the measured and calculated critical time and temperature for each test. In general, the agreement is good except for a few cases (such as Test 14). The discrepancy is possibly resulted from the assumed steel strength degradation at elevated temperatures due to

lack of material test data. Figure 16 plots the temperature and the mid-span deflection predictions against experimental results for test 11. 8.3. Two-dimensional Frames Two frames (Li et al. 1997; Zhao 1995) have been studied. The configuration of the frame and the loading

Advances in Structural Engineering Vol. 7 No. 4 2004

323

Performance Based Fire Safety Design of Structures A Multi-dimensional Integration

Table 1. Critical temperature of uniformly heated members Simply Supported Beam in Bending 4-side BS5950 heated Analysis % difference Critical Temperature (C) at load ratio R1 0.2 715 725 1.4 0.3 660 671 1.7 0.4 620 629 1.5 0.5 585 591 1.0 0.6 555 559 0.7 0.7 520 527 1.3

Critical Temperature (C) at load ratio R2 Column in Compression 4-side BS5950 heated Analysis 70 % difference 4-side BS5950 heated Analysis >70 % difference
R1 = Mf /Mc : slenderness ratio Mc: moment capacity at ambient temperature

0.2 710 723 1.9 635 662 4.2

0.3 655 678 3.5 635 649 2.2

0.4 615 641 4.2 590 611 3.5

0.5 580 608 4.8 545 568 4.2

0.6 540 564 4.5 510 529 3.8

0.7 510 529 3.8 460 478 3.9

R2 = F/Agpy + Mx/Mcx + My/Mcy Mf: applied mid-span moment at fire F: axial force Ag: cross-section area py: yield strength Mx and My: applied major and minor axis moment at fire Mcx and Mcy: major and minor axis moment capacity at ambient temperature

Table 2. Critical time and temperature of UK standard fire tests Critical Time (min) Test No Test 2 Test 3 Test 4 Test 5 Test 6 Test 7 Test 8 Test 9 Test 10 Test 11 Test 12 Test 13 Test 14 Test 89 Test 90 Test 91 Test 92 Test 93 Load Ratio 0.50 0.57 0.36 0.61 0.37 0.36 0.36 0.37 0.49 0.50 0.53 0.40 0.25 0.50 0.65 0.34 0.05 0.09 Test 22.5 22.0 29.0 26.7 22.8 22.3 21.3 24.2 20.5 21.4 28.4 25.1 26.4 20.0 20.7 23.0 117.0 75.0 Analysis 20.7 20.6 28.2 22.9 25.9 24.2 24.4 26.4 21.0 20.8 29.3 24.3 33.2 22.4 19.0 29.5 109.0 56.4 Error % 8.0 6.4 2.8 14.2 13.6 8.5 14.6 9.1 2.4 2.8 3.2 3.2 25.8 12.0 8.2 28.3 6.8 24.8 Test 660 634 701 647 737 731 705 714 655 683 681 727 745 651 630 705 1061 977 Critical Temperature (C) Analysis 693 682 745 694 734 743 742 734 709 706 680 736 791 692 640 742 1046 932 Error % 5.0 7.6 6.3 7.3 0.4 1.6 5.2 2.8 8.2 3.4 0.1 1.2 6.2 6.3 1.6 5.2 1.4 4.6

is shown in Figures 17 and 18. In both studies, the measured temperature of each member is used in the structural analysis. Figures 19 and 20 illustrate the excellent correlation between the predicted and the measured displacements.

8.4. Three-dimensional Frame A full scale fire test was carried out on a threedimensional steel tubular frame in SINTEF. The test details and the verification procedures can be found in Skallerud and Amdahl (2002). The proposed advanced

324

Advances in Structural Engineering Vol. 7 No. 4 2004

J. Y. Richard Liew

0.20

800

0.16

600

0.12 400 0.08

Test (deflection) Analysis (deflection)


200

0.04

Test (temperature) Analysis (temperature)


0 0 10 20 30 40 50

0.00

Time (min) Figure 16. Test and analysis results for test 11
30 30 30 30 30 30 30 kN 82 kN 1400 mm A B C 82 kN D 9.45 kN

540 540 540 540 540 540 7.2 100 4.2 100 column cross section 100

1500 mm

1500 mm 6.0 100 56 beam & column cross-section Horizontal displacement at D


10

4.5 55

4.5

beam cross section

Figure 17. Lis frame

Figure 18. Zhaos frame

Horizontal displacement (mm)

20 Displacement (mm) Node B, Analysis Node B, Test 0 Node A, Analysis 0 10 Node A, Test 20 30
20

Analysis Test

10

Vertical displacement at C

10

0 0 10 20 30

Time (min) Figure 19. Analysis and test results (Lis frame)

Time (min)

Figure 20. Analysis and test results (Zhaos frame)

analysis is used for the verification. It has been found that the correlation between simulation and test results with respect to the mechanical response was extremely good. The primary collapse load obtained was almost perfectly predicted.

9. FIRE ANALYSIS OF AN OFFICE BUILDING


This section presents a performance-based approach for analysing a six-storey building frame subject to various scenarios of fire attack. Advanced analysis method (Ma

Advances in Structural Engineering Vol. 7 No. 4 2004

Lower Flange Temperature

Mid-Span Deflection (m)

325

Performance Based Fire Safety Design of Structures A Multi-dimensional Integration

W10 60

Y W12 26 W12 26

H = 6 x 3.658 m = 21.948 m

W12 87

W12 53

W12 53

W12 120

W10 60

X W12 26 7.315 m 1 2 W12 26 7.315 m 3

W12 87 X 3

7.315 m

W12x87

PLAN

FRONT ELEVATION

Figure 21. Six-storey building frame

and Liew 2004) is used to propose a much reduced fire protection plan for beams and columns and prove that the design can satisfy the performance criteria of fire safety. The building is classified as office building as shown in Figure 21. 9.1. Limit States Design The frame is designed for the strength limit state at ambient temperature according to Eurocode 3 with the following actions: 3.6 kN/m2 Permanent Dead load (Gk,1) actions Gk: Permanent imposed 1.9 kN/m2 load (Gk,2) Variable Variable imposed 1.6 kN/m2 actions Qk: load (Qk,1) 593 kN (in Wind load (Qk,2) Y-direction) The beam and column sizes are indicated in Figure 21. A36 steel (design strength of 250 N/mm2) is used in all sections. Plastic hinge analysis method is adopted. Each beam is modelled using 4 elements and each column using 1 element. Wind load is simulated by applying a point load in Y-direction at every beamcolumn joint of the front elevation. At the fire limit state, which is treated as accidental loads in Eurocode 3 Part 1-2 (2001), the design effect of the actions is expressed as: Efi,d,t = Gk + 1Qk,1 + 2Qk,2 (15)

Where 1, 2 are factors due to the probability of loads acting individually or in combination. Depending upon which variable load is the dominant action, two load combinations are possible under fire limit state: Load combination 1: Efi,d,t = Gk,1 + Gk,2 + 0.5Qk,1 + NL (Notional Load) Load combination 2: Efi,d,t = Gk,1 + Gk,2 + 0.3Qk,1 + 0.5Qk,2 In load combination 1, the notional load is taken as 0.5% of the factored gravity load at each storey, applied in Y-direction and is distributed to the beam-column joints as point load. In both cases, the structure is subjected to gravity load or the combination of gravity load and wind load first, followed by fire. 9.2. Fire Modelling Parametric fire recommended in Eurocode 1 Part 1-2 (2001), is used to simulate the fire in the compartment by considering the type of building, floor layout, realistic fire load and possible fire fighting measures. Fire load density per floor area qf,k = 420 MJ/m2 is adopted for common office building. The design fire load qf,d is defined as: qf,d = qf,k . m . q1. q2 . n (16)

where m is the combustion factor and is assumed as 0.8; q1 is the partial factor taking into account the fire activation risk due to the size of the compartment. For floor area from 25 m2 up to 250 m2, q1 is equal to

326

Advances in Structural Engineering Vol. 7 No. 4 2004

J. Y. Richard Liew

1.5 (in this case the floor area Af is 53.5 m2). q2 is 1.0 for occupancies such as office, residence and hotel. It is assumed that no automatic fire suppression and detection system is installed but an off site fire brigade is available from which n is calculated as 0.78. The design fire load qf,d is thus computed as 393 MJ/m2, which is equivalent to 98 MJ/m2 per total area (qt,d). The surrounding surfaces of the compartment are assumed to
Temperature (C)
1200

be normal concrete with b value of 1900 J/m2s1/2K. Assuming an opening factor OF = 0.4, the temperature time curve is plotted as shown in Figure 22. It can be seen from Figure 22 that the fire curve with opening factor OF = 0.04 are below the standard ISO 834 fire curve, providing the possibility of reducing passive fire protections. Two compartments are considered as shown in Figure 23. The columns in compartment 1 are

1000

ISO 834

800

OF = 0.04
600

400

200

0 0 20 40 60 80 100 120

Time (Min) Figure 22. Parametric time-temperature curves for six-storey office building frame

42 43 40 41 Element number 4 5 1 Z Y X 7

Fire compartment 2

Fire compartment 1

Figure 23. Fire compartments in 6-storey space frame

Advances in Structural Engineering Vol. 7 No. 4 2004

327

Performance Based Fire Safety Design of Structures A Multi-dimensional Integration

Buckling of Column 4

Z Y X

Load Combination 1 (no collapse)

Load Combination 2

Figure 24. Deformed shapes of 6-storey frame for load combinations 1 and 2

0.40

Column Head Deflection (m)

0.35 0.30 0.25 0.20 0.15 0.10 0.05 0.00 0

Case 1: Column Unprotected Case 2: Column Protected

4 1
5 10 15 20 25 30 35

Time (min) Figure 25. Column 4 head displacement in the Y-direction

most heavily loaded and the size of columns reduces from the 4th storey onwards. 9.3. Fire at the First Storey Compartment If all the beams and columns in the lower floor fire compartment are unprotected, it is found that load combination 2 is more severe as the structure fails at a critical time of 33.7 minutes while it can survive the fire under load combination 1. The deformed shape of the structure under fire for each load combination is shown in Figure 24. Under load combination 1, as the beams expand under fire, column heads are forced to open up in both X and Y directions. When the frame is subjected to load combination 2, the effect of wind load in Y-direction becomes pronounced, causing the frame to deform in a twisting mode. Despite the expansion of the heated

beams, all columns (1, 2, 4 and 5) sway to the same direction as the wind load. The center of gravity of the frame thus shifts to the leeward columns (4 and 5), producing large axial force in the columns (Figure 25). It is the failure of column 4 which triggers the collapse of the frame under fire. Load combination 2 is found to be most critical; therefore subsequent analyses are carried out using only this load combination. Since the columns are found to be the critical members, it is proposed that all the fire affected columns are fully protected. Second-order plastic hinge analysis is again carried out on the partly protected structure, and the displacement of the column head in Ydirection is found to be greatly reduced (see Figure 25), in contrast to the runaway deflection of column 4 when approaching failure for the unprotected columns.

328

Advances in Structural Engineering Vol. 7 No. 4 2004

J. Y. Richard Liew

Time (min)
0 0
Beam 9

10

15

20

25

30

35

Beam Mid-Span Deflection (m)

0.05 0.1 0.15

9
0.2 0.25
Case 1: Column Unprotected Beam 7 Case 2: Column Protected

Figure 26. Mid-span deflection of Beams 7 and 9 Time (min)


0 0.35 0.30 0.25 0.20 0.15 0.10 0.05 0.00 0.05 0.10 0.15 Column 40 Column 40 Column 42 5 10 15 20 25 30 35 Deflection in Y-direction Deflection in X-direction 42 40

Column Head Deflection (m)

Column 42

Figure 27. Column head displacements in X and Y directions (unprotected)

However, the critical time shows only marginal improvement, from 33.7 min to 35.5 min. Beam collapse mechanism forms at beams 7 and 9 (the smallest beams) in X-direction due to large restraint force from the supporting columns, causing larger mid-span deflection as shown in Figure 26. The next logical step is to provide fire protection to beams 7 and 9. In this case, the building survives the entire fire duration with plastic hinges occur at the beams in Y-direction. 9.4. Fire at the Fourth Storey Compartment Fire is assumed to occur at the 4th storeys compartment as shown in Figure 23. When all the members are unprotected, extensive plastic hinges form at the four columns, triggering the collapse of the frame during the fire. Although the loads on the fourth storeys columns are smaller than those on the first storeys columns, the column size from the fourth storey onward is also smaller.

At a critical time of 33 min, columns experience runaway deflections in both X and Y directions (Figure 27), symbolizing the failure of the columns. Figure 28 shows the axial force in windward and leeward columns. The shift of the center of gravity due to wind load produces larger axial force in the leeward column. If a realistic fire model is considered, it is possible to reduce the cost of fire protection. In this 6-storey building frame, columns and beams in X-direction require passive fire protection from 1st storey to 3rd storey. But from 3rd storey onwards, only columns need to be protected while all the beams can be left unprotected. However if ISO standard curve is used irrespective of layout of the building, fire loads and ventilation condition, all the members in the building need to be fire protected Further study will be carried out on high-rise buildings where the savings in passive fire protection may become more significant if a realistic fire is considered.

Advances in Structural Engineering Vol. 7 No. 4 2004

329

Performance Based Fire Safety Design of Structures A Multi-dimensional Integration

Time (min)
0 50 100 150 200 250 300 Column 42 40 Column 40 42 0 5 10 15 20 25 30 35

Column Axial Force (kN)

Figure 28. Axial force in columns 40 and 42 (unprotected members)

10. FIRE PROTECTION


Fire protection falls into two main categories, prevention and protection. Preventative measures include control of flammable inventories, control of ignition sources, monitoring of environmental conditions leading to initiation of alarms and automatic process control, and fire detection systems designed to extinguish fires immediately on detection. Protection measures fall into two categories, passive and active. Passive measures include fire barriers, fire resistant enclosures, fire doors, fire retardant coatings and fire protective coatings. Active measures include water and chemical sprays or deluges, foam dispersion and inert gas dispersal. After the WTC incident, people expect the public buildings and their work place to be designed to allow safe evacuation in the event of fire or explosion. Certain industrial buildings have stringent fire protection requirements. For examples, oil and gas production and processing, nuclear related product storage and processing, chemical process and storage and key infrastructures and transportation routes are facilities that attracted greater risks. From safety and licensing authorities point of view, the structure must be capable of safe evacuation in the event of fire. From the owners view point, the fire protection is often an expensive statutory feature, which requires initial capital investment. From the operator point of view, fire protection must be maintained to preserve the safety margins declared in the safety documentation. The volume and cost of passive fire protection materials are often a critical factor for consideration; therefore, there is a need to balance these conflicting requirements when specifying fire protection. Fire protection can be provided as an all encompassing scheme, or it can be functionally designed

to optimize on cost, weight and maintenance. Research has shown that fire can be successfully suppressed using a properly design and maintained active protection system without the need of passive fire protection. Evidently the structural member affected by the fire may not be reusable, but there is no guarantee that a fire protected structure could be re-used anyway. Computer models have been developed to predict the response of structures considering fire protection materials. This model will include the basic thermal transmission phenomena, radiation, conduction and convection, the temperature dependent properties of materials and the location and nature of fire protection measures. The resulting thermal histories are then applied to the structure to predict time to collapse, or to demonstrate the degree of collapse. From this, the structure can be economically protected to meet the safety requirements.

11. INTEGRATED EXPLOSION AND FIRE ASSESSMENT OF STRUCTURES


The assessment of the response of structures to explosion is an increasingly important factor in design, particularly where the storage and processing of explosive materials is concerned. Many structures are required to be blast resistant to protect personnel and adjacent facilities, and to reduce the possibility of escalation of events. These structures are therefore designed to contain the effects of explosion or to act as a significant barrier. There is a great difference in the structural behaviour of buildings subject to explosion and fire loads. The short duration of explosion loading implies that the material is strain-rates dependent, i.e., high strain rate will increase the yield strength of steel (Izzuddin & Fang 1997). On

330

Advances in Structural Engineering Vol. 7 No. 4 2004

J. Y. Richard Liew

the other hand, fire loading is associated with elevated temperatures which cause thermal strains and lead to significant deterioration in the material properties of steel. Izzuddin et al. (2000) proposed an integrated analysis method for explosion and fire analysis of steel structures incorporating material models which account both for rate-dependency and thermal property of steel. Liew and Chen (2004) presented a numerical approach for inelastic transient analysis of steel frame structures subjected to explosion loads followed by fire. The proposed transient inelastic analysis can be used effectively to solve explosion and thermal response problems, taking account of geometric and material nonlinearities. To achieve both computational accuracy and efficiency, an analysis procedure has been proposed in which fire can be treated as a separate event after the occurrence of an explosion. The fire resistance of the structure can be evaluated by analyzing the deformed geometry of the structure caused by the explosion. The approximate fire analysis method does not require time domain solutions but it predicts higher fire resistance than the strict inelastic transient analysis. Nevertheless, it offers an alternate means to evaluate the performance of structures subjected to combined scenarios of explosion and fire at a much reduced computational cost. Hand calculation procedure and computational techniques have been developed with the aim of predicting how a structure will respond under the interaction of blast and fire. The followings summarized some of the works that are being investigated using the explosion response technologies developed. Pseudo-dynamic or pseudo-static methods these are the simplest to apply since they take the explosion overpressure as a blanket loading, and are usually combined with dynamic amplification factors. The methods can assess both elastic and inelastic responses and can be applied to complex structures. Single Degree of Freedom Method this is a dynamic analysis technique, which predicts the response of a structure by reducing the structure to a simplified spring/mass system. The method is effective for simple structures that behave in a similar manner as to a spring/mass system. The method can assess both linear and non-linear responses and can be applied to simulate the response behaviour of more complex structures. Finite Element Analysis they can be used effectively to solve explosion response problems, taking account of geometric and material nonlinearities. In term of computation time usage, balance has to be sought since there is a fine line between a model which is detailed enough to

predict the response, and coarse enough not to run for impracticable lengths of time. Evidently, with the continued increase in the speed and capacity of computers, this problem will be less critical. The model detail is a matter to be addressed by the analysts carrying out the work as is the choice of solution method. The two most common time domain solutions, implicit integration and explicit integration can both be used, each having their own pros and cons. Significant experience has been gained in the assessment of structures subjected to explosion loading using both hand and FEA based methods. Combining the structural response analysis with the explosion prediction analysis, it is possible to predict and hence optimize structural resistance.

12. CONCLUSIONS
The difference in perspective between architect and engineer is noticeable in the design and construction process. In the conventional approach, the architect would specify the fire designs based on prescriptive code requirements and the engineer would design the structures with fire protection to achieve a certain fire rating. It becomes apparent in the recent years that the structural engineers should directly involve in fire engineering rather than the traditional approach of the architect specifying the fire designs. In a performance based design approach, the first step is to understand what level of performance is expected, then to design to these levels and followed by predicting the performance that will be achieved, and finally to be able to assure the reliability and robustness of the design in the occurrence of an extreme event. To carry out a quantitative assessment on the performance of a building in fire requires the knowledge of fire science, material properties at elevated temperature, occupant behaviour and evacuation procedure during an emergency situation, heat transient and structural response phenomena, and fire protection etc. All these would require a multi-dimensional integration approach, as described in this paper, for performance-based design of structures. Fire may be treated as a building load, consistent with the treatment of other loads in building design such that it can be integrated with structural design in various load combinations. Design fire scenarios can be prescribed for standard building forms and further examined for more complex systems. Structural design should also consider the integration between evacuees and firefighter interactions. Proper fire model with interaction with the active and passive protection measures should be developed, and relationship between emergency

Advances in Structural Engineering Vol. 7 No. 4 2004

331

Performance Based Fire Safety Design of Structures A Multi-dimensional Integration

response and fire resistance should be established so that appropriate performance-based method can be developed to predict the building response to extreme events. The examples given in this paper illustrated that saving in passive fire protection could be substantial for high-rise and large building framework if realistic natural fires, instead of the conventional standard fire, were considered in design. Improved understanding of the real behaviour of natural fire in tall buildings opens new ways of integrating fire safety and structural design. Prescriptive codes without considering the systems limit states behaviour, are often quite approximate in nature. With the advance in computing technologies, there is an increasing demand for robust and efficient nonlinear analysis methods for performance-based design of structures subject to fire and explosion. Some of the works mentioned in this paper are a step towards this development.

ACKOWLEDGEMENTS
The author would like to acknowledge the contributions made by Dr H Chen, Dr L K Tang, Ms K Y Ma, and Ms H X Yu for their research work on steel structures in fire in the Department of Civil Engineering at the National University of Singapore. The work is funded by research grants (R264000138112) made available by the National University of Singapore.

REFERENCES
ECCS (1993). Fire Safety in Open Car Parks, Modern Fire Engineering, Technical Committee 3, European Convention for Constructional Steelwork, Brussels, Belgium. Eurocode 1: Part 1-2 (2001). Actions on Structures Part 1-2: General Actions Actins on Structures Exposed to Fire, EN 1991-1-2, European Committee for Standardization, Brussels. Eurocode 3: Part 1-2 (2001). Design of Steel Structures Part 1-2: General Rules Structural Fire Design, prEN 1993-1-2, European Committee for Standardization, Brussels. Feasey, R. and Buchanan, A. (2002). Post-flashover fires for structural design, Fire Safety Journal, Vol. 37, pp. 83-105. Franssen, J. M. (1997). Improvement of the parametric fire of eurocode 1 based on experimental test results, Fire Safety Science Proceedings of the Sixth International Symposium, pp. 927-938. Izzuddin, B. A. and Fang, Q. (1997). Rate-sensitive analysis of framed structures - Part I: model formulation and verification, Structural Engineering and Mechanics, Vol. 5, No. 3, pp. 221-237. Izzuddin, B. A., Song, L., Elnashai, A. S. and Dowling, P. J. (2000). An integrated adaptive environment for fire and explosion analysis of steel frames - Part II: verification and application, Journal of Constructional Steel Research, Vol. 53, No. 1, pp. 87-111.

Li, G. Q., Jiang, S. C., Lin, G. X. and Yu, L. H. (1997). An experimental research on behaviour of steel frames subjected to fires, Proceedings of the 3rd Sino-Japan Conference on Structural Technology, 4-8 November, Shenzhen, P. R. China, pp. 68-78. Liew, J. Y. R., Tang, L. K., Holmaas, T. and Choo, Y. S. (1998). Advanced analysis for the assessment of steel frames in fire, Journal of Constructional Steel Research, Elsevier Science, UK, Vol. 47, No. 1-2, pp. 19-45. Liew, J. Y. R., Chen, H., Shanmugam, N. E. and Chen, W. F. (2000). Improved nonlinear plastic hinge analysis of space frames, Engineering Structures, Vol. 22, pp. 1324-1338. Liew, J. Y. R. and Tang, L. K. (2000). Advanced plastic-hinge analysis for the design of tubular space frames, Engineering Structures, Elsevier Science, UK, Vol. 22, No. 7, pp. 769-783. Liew, J. Y. R. ed (2002.). Structures in Fire, Proceedings of Int. Sym. on Structures in Fire, Oct. 25, 2002, Singapore, Singapore Structural Steel Society. Liew, J. Y. R., Tang, L. K., and Choo, Y. S. (2002). Advanced analysis for performance-based design of steel structures exposed to fires, Journal of Structural Engineering, ASCE, USA, Vol. 128, No. 12, pp. 1584-1593. Liew, J. Y. R. and Chen, H. (2004). Explosion and fire analysis of steel frame structures using fiber element approach, Journal Structural Engineering, ASCE, July Issue, In Press. Ma K. Y. and Liew J. Y. R. (2004). Nonlinear Plastic Hinge Analysis of Three-dimensional Steel Frames in Fire, Journal Structural Engineering, ASCE, July Issue, In Press. Moss, P. J. ed (2002). Structures in Fire, Proceedings of Second International Workshop on Structure in Fire, March 18-19, Christchurch, New Zealand. Schleich, J. B. (1996). A natural fire safety concept for buildings 1, Fire, Static and Dynamic Tests of Building Structures, E & FN SPON. SINTEF (1995). Fire and Heat Transfer Simulations (FAHTS) of Frame Structures, Theory and Users Manual, SINTEF Structures and Concrete, Trondheim, Norway. SFPE (2002). Handbook of Fire Protection Engineering, 3rd Edition, P.J. DiNenno, (Ed.), NFPA: Quincy, MA. Skallerud, B. and Amdahl, J. (2002). Nonlinear Analysis of Offshore Structures, Research Studies Press, England, 323p. Wang, Y. C., Lennon, T. and Moore, D. B. (1995). The behaviour of steel frames subject to fire, Journal of Constructional Steel Research, Elsevier Science, UK, Vol. 35, pp. 291-322. Wainman, D. E. and Kirby, B. R. (1988). Compendium of UK Standard Fire Test Data: Unprotected Structural Steel 1& 2, British Steel Corporation, Ref. No. RS/RSC/S10328/1/87/B. Yao, Z. P., Wang, R. J. and Zhang, X. J. (1995). Heat Transmission, Beijing Science and Technology University, Beijing, P. R. China. Zhao, J. C. (1995). Fire Resistance of Steel Framed Structures, Ph. D. thesis, Tong Ji University, Shanghai, P. R. China.

332

Advances in Structural Engineering Vol. 7 No. 4 2004

J. Y. Richard Liew

Jat-Yuen Richard Liew is an associate professor and the director of the centre for construction materials and technology in the National University of Singapore (NUS). He received his B.Eng and M.Eng degrees (Civil Engineering) from NUS and his Ph.D degree from Purdue University in 1992. His research interests include deployable structures, steelconcrete composite systems and fire safety design of buildings. Arising from these works, he has generated some 150 technical publications. These include technical articles in journals, presentations, reports, books, and patents. He interacts closely with the steel industry in the Asian region as a technical advisor in the areas of steel and composite structures. He has also seen his R&D brought from the laboratory to full-scale applications. The latter include projects in airport structures, high-rise buildings, large-span and prestressed structures. He is a registered professional engineer in Singapore and a chartered engineer in U.K.

Advances in Structural Engineering Vol. 7 No. 4 2004

333

You might also like