You are on page 1of 11

Engineering Structures 24 (2002) 10151025 www.elsevier.

com/locate/engstruct

Static, seismic and stability analyses of a prototype wind turbine steel tower
N. Bazeos, G.D. Hatzigeorgiou, I.D. Hondros, H. Karamaneas, D.L. Karabalis , D.E. Beskos
Department of Civil Engineering, University of Patras, 26500 Patras, Greece Received 11 December 2000; received in revised form 7 February 2002; accepted 8 February 2002

Abstract Selected results of a study concerning the load bearing capacity and the seismic behavior of a prototype steel tower for a 450 kW wind turbine with a horizontal power transmission axle are presented. The main load bearing structure of the steel tower rises to almost 38 m high and consists of thin-wall cylindrical and conical parts, of varying diameters and wall thicknesses, which are linked together by bolted circular rings. The behavior and the load capacity of the structure have been studied with the aid of a rened nite element and other simplied models recommended by appropriate building codes. The structure is analyzed for static and seismic loads representing the effects of gravity, the operational and survival aerodynamic conditions, and possible site-dependent seismic motions. Comparative studies have been performed on the results of the above analyses and some useful conclusions are drawn pertaining to the effectiveness and accuracy of the various models used in this work. 2002 Published by Elsevier Science Ltd.
Keywords: Wind turbine; Steel tower; Static analysis; Seismic analysis; Stability analysis

1. Introduction This work depicts some critical aspects of the analyses performed during the design of an almost 38 m high steel tower supporting a prototype 450 kW wind turbine with a horizontal power transmission axle. The entire wind turbine system, now under construction, is the rst designed and manufactured exclusively in Greece. The main supporting structure of the wind turbine, as shown in Fig. 1(a), is assembled by thin-wall cylindrical and conical parts of varying diameters and wall thicknesses. Circular stiffeners are placed at regular intervals along the height of the structure for further stiffening against local buckling. The optimal vertical spacing between circular stiffeners with regard to the thickness and the diameter of the shell structure has been the subject of an extensive stability analysis study. Furthermore, at the base of the structure a substantial door opening is considered. The adverse inuence of this

Corresponding author. Tel.: +30-610-996218; fax: +30-61997812. E-mail address: karabali@upatras.gr (D.L. Karabalis).

opening on the overall structural behavior of the tower is partly counterbalanced by heavy reinforcement along its perimeter as shown in Fig. 2. However, to the best of our knowledge there are no formal analytical procedures for estimating the effectiveness of such design congurations. It is also considered common practice to analyze such tower-like structures as xed at their base with no consideration of the foundationsoil interaction. In most cases, over-designed foundations and stiff soil conditions could, for all practical purposes, produce such a xed base. However, this is not always the case since in many instances wind turbines are installed in regions with relatively soft soil deposits. Under these circumstances the interaction of the structure with the supporting soil, particularly under external dynamic or seismic loads, could become a major concern in designing the foundation and subsequently the entire structure of the wind turbine tower. On the basis of the previous discussion, three major aspects of the particular design/analysis of the tower structure and its foundation are primarily discussed in the following work:

0141-0296/02/$ - see front matter 2002 Published by Elsevier Science Ltd. PII: S 0 1 4 1 - 0 2 9 6 ( 0 2 ) 0 0 0 2 1 - 4

1016

N. Bazeos et al. / Engineering Structures 24 (2002) 10151025

Fig. 1.

(a) Steel tower, (b) ange connections, (c) circular stiffener and intermediate landing, and (d) intermediate landing.

N. Bazeos et al. / Engineering Structures 24 (2002) 10151025


Cover plate 16 2100*1300 St52 3N 14 welding openings 100*30(a=8mm) 130
16*500/140*50 0

1017

200

16*5 00/14 0*50

220

16 *35 0/1 40 *50

180

30

16*220/120*50 *
450

15

400

22.5

348 350 350


50 r=2

490

16

610

0/

12

0*

280

50

r=4 50

25

342 250

45

30

Fig. 2. Details of door opening geometry, discretization and representative results of static analysis (stresses in Mpa).

Fig. 2.

170

16 *

Continued

1500

700 0

810

350

1018

N. Bazeos et al. / Engineering Structures 24 (2002) 10151025

1. The accurate and efcient determination of the minimum load combination that would produce local buckling phenomena on the shell structure. The computation of this load combination is directly related to a number of geometric variables associated with the arrangement of the various load bearing elements of the structure. 2. The inuence of the door opening, and of the array of stiffeners used around it, on the overall structural behavior of the tower structure. 3. The accurate and efcient determination of the dynamic characteristics and the possible seismic behavior of the tower structure: (a) on a xed base, and (b) including dynamic soilstructure interaction (SSI) effects. The main assumption underlying most of the analyses considered in this work is that under any load combination (including any appropriate load safety factors) the material of the load bearing structural elements of the tower should remain in the linear elastic region of its stressstrain diagram. Material linearity is an essential prerequisite of all the pertinent building regulations, e.g. Germanischer Lloyd [1], in view of the expectation that wind turbines should survive and remain operational, during their entire life cycle (usually 20 years), against fatigue damage which remains the governing cause of failure for the tower structure [2]. Low operational stresses provide some cushion against fatigue damage [3]. The static, stability and dynamic analyses performed in this work are based on a number of rened nite element and other simplied models. The results of these analyses are used to check the load bearing capacity of the structure against the recommendations of widely accepted design codes, mainly the Eurocode 3 [4] and the German Standards [5]. Due to lack of space, only a few characteristic examples of these analyses are shown in the next sections. For a complete list of analyses and results the reader is referred to Beskos et al. [6] and Karabalis et al. [7].

56 at the base ange connecting to the foundation to 64 at elevation +13.64 m and 40 at elevation +26.35 m). Circular ring stiffeners are placed at regular intervals along the height of the structure, as shown in Fig. 1(c). These stiffeners and a number of intermediate landings, as shown in Fig. 1(d), provide additional resistance against local buckling by reducing the free height between end-anges. Steel St52, with yield stress fy,k 360N / mm2, is used for all the load bearing structural elements. Two sets of static loads are considered: (a) The pseudostatic aerodynamic loads under survival and operational conditions suggested by Riziotis and Voutsinas [8]. The concentrated aerodynamic loads at the elevation of the power transmission axis (el. +38 m), listed in Table 1, are due to the wind resistance and/or operation of the runner. In addition, aerodynamic loads distributed along the body of the tower itself have been computed by Riziotis and Voutsinas [8] and are accounted for in the analysis under survival conditions. Aerodynamic loads under survival, shut down, conditions have a recurrence period of 50 years, and assume that the entire system is shut down with the runner securely immobile. The safety margin of the load carrying capacity of the structure under survival conditions should not be less than one, after, of course, the application of the appropriate safety factors to the loads and the structure. The anticipated loads under operational conditions, is used for the computation of stresses necessary in fatigue analysis. Wind turbine structures are usually designed to develop low stress levels at operational aerodynamic loads, in an effort to expand their productive life span. (b) The second set of static loads due to gravity, consists of a concentrated load at the top of the tower representing the weight of the nacelle, runner, generator, gear box, etc. (236 kN with 0.75 m
Table 1 Pseudo-aerodynamic loads at elevation +38 m (Riziotis and Voutsinas [8]) Operational Survival conditions (wind conditions (wind velocity 25 m/s) velocity 70 m/s) Fx (kN) Fy (kN) Fz (kN) Mx (kN m) My (kN m) Mz (kN m) 75.87 0.23 226.25 161.29 95.40 4.88 216.23 0.28 0 96.03 82.92 5.38

2. Description of structure and FE model The 36,984 m vertical steel tower, as shown in Fig. 1(a), is assembled by thin-wall cylindrical and conical pieces, of varying diameters (2800 mm at the base to 1820 mm at the top) and wall thicknesses (16 mm at the base to 10 mm at the top), which are welded together along their perimeters. However, for transportation and erection purposes the entire tower is pre-assembled into three sections which are bolted together by means of heavy circular end-anges (each 70 mm thickness), as shown in Fig. 1(b), and arrays of post-tensioned bolts along their perimeter (the number of bolts varies from

Note: The x-axis is horizontal in the direction of the wind, while the z-axis is vertical with an upward direction.

N. Bazeos et al. / Engineering Structures 24 (2002) 10151025

1019

Fig. 3.

Samples of the results obtained from the static analysis of the entire tower (stresses in Mpa, displacements in mm).

1020

N. Bazeos et al. / Engineering Structures 24 (2002) 10151025

eccentricity along the x-axis) and the weight of the tower itself distributed along its height (78,500 N/m3, specic weight of steel). The safety factors for the static loads are specied as [1]: Favorable gravity loads: 1.00 Unfavorable gravity loads: 1.35 Aerodynamic loads: 1.50 The seismic loads are in accordance with the specications of the Greek Seismic Code [9] where the design seismic motion has a 10% likelihood of being exceeded during a period of 50 years. The corresponding elastic design spectrum of horizontal acceleration Re(T) is dened as: Re(T) Re(T) Re(T) Ag1 1 T (hb0 1) , 0 T T1 T1 (1)

mm thick ring end-anges, see Fig. 1(b). The intermediate landings, see Fig. 1(c) and (d), are simulated by three-dimensional (3-D) beam elements. The whole structure is divided into 3763 elements with 12,539 nodes and a total 69,186 degrees-of-freedom.

3. Numerical results and discussion 3.1. Static analysis Pseudo-aerodynamic (survival conditions) and gravity loads are considered in static analysis. The maximum shear stresses, according to the Tresca failure criterion, appear in two small regions, in close proximity to either side of the door opening and their values do not exceed 100 N/mm2. The maximum von Mises stresses also

Ag1hb0, T1 T T2 T2 Ag1hb0 , T2 T T

where T=period in seconds; T1, T2=characteristic cut-off periods for different soil conditions (for rock and semirock conditions, as in this case, T1=0.1 s, T2=0.40 s); A=site specic maximum acceleration (in this case 0.12 g, for the city of Lavrio or the South Evia region); g1=signicance factor (in this case g1=1.0, usual occupational signicance); h=correction factor for damping 7 / (2 x)); b0 2.5 ratios x other than 5% (h (design spectra multiplier). According to the Greek Seismic Code [9] a total of six statistically independent articial seismic motions enveloping the above design spectrum should be used if the analysis is performed in time domain. For the purposes of this study, articial accelerograms compatible with the Greek Seismic Code [9] are developed with the aid of the computer code SRP [10]. Subsequently, the seismic analysis of the tower is carried out in direct time domain. In addition, the inuence of the compliance of the supporting soil medium upon the dynamic characteristics of the tower, as compared to the xed base analyses, is calculated as discussed in Section 3. The static, stability and seismic analyses presented in this work have been carried out using a rened nite element model in conjunction with the general-purpose program NISA II [11]. Fixed boundary conditions are assumed at the lower end of the tower for all static analyses. Quadrilateral shell elements (eight nodes per element) are used for the tower wall, the trapezoidal vertical stiffeners adjacent to the ange connections and the L-shaped ring stiffeners, see Fig. 1(b). Hexahedral solid elements (20 nodes per element) are adopted for the 70

Fig. 4. Distribution of normal stresses and expected buckling shape (stresses in Mpa).

N. Bazeos et al. / Engineering Structures 24 (2002) 10151025

1021

appear in two small regions on either side of the door opening and their values do not exceed 211 MPa. The maximum shear and von Mises stresses appear in the above regions regardless of the directionality of the aerodynamic loads. The geometry of the door-opening region, the arrangement of the surrounding stiffening elements and representative results of static analyses are shown in Fig. 2. The effectiveness of the arrangement of the stiffening elements around the door opening in keeping the stresses at acceptable levels is evident. The largest von Mises stress in the cylindrical or conical thin-wall parts appear at 13,237 mm above the base when the directionality of the aerodynamic loads is such that compressive stresses appear over the region diametrically opposite to the door opening. As expected, the maximum horizontal displacement umax 362mm, less than approximately one-hundredth of the total height, appears at the top of the tower. Characteristic samples of the results obtained from the static analyses of the entire tower appear in Fig. 3. 3.2. Buckling analysis An elastic buckling analysis is performed in order to investigate the critical coupled stress state due to axial, exural and torsional loads. The purpose of this eigenvalue buckling analysis is limited to the identication of the bifurcation points on the primary loaddeection path of the structure. This requires an eigenvalue extraction using the governing equation: (K liKg)ui 0 (2)

indrical and conical portions and appropriate edge conditions. Thus, in this work, the critical buckling load of all the cylindrical and conical parts between stiffeners is computed according to DIN 18800 [5]. The intermediate stiffening elements are substituted by appropriate boundary conditions, i.e. depending on the type of the stiffening element the radial expansion and/or the relative rotation among adjacent portions are restrained. The relative difference between the results obtained by the nite element based analyses and the procedures recommended by DIN 18800 [5] does not exceed 1.2% [7]. This agreement proves the accuracy of the nite element model used in this work and the relative effectiveness of the simplied procedures recommended by DIN 18800 [5]. The above, nite element based, geometrically nonlinear analyses with incremental loading, can produce

18 17 16

36.984 35.637 34.891

Altitude (m)

15

31.051

14 13 12 11 t= 10mm 10 9 8 7 6 5 4

26.251 25.650 23.204 21.430 18.204 17.350 13.637 12.550 10.570 7.870 7.070

where K is the linear stiffness matrix, Kg the geometric or initial stress stiffness matrix, li the ith eigenvalue and ui is the corresponding modal shape. In an effort to achieve an as high as possible level of convergence, more than 15 nite element models have been created for the generation of the stiffness and geometric stiffness matrices. In the beginning, a rough nite element mesh is created and the critical regions are identied. Successive mesh renements of the critical regions make possible the precise computation of the local buckling effects. The minimum static load multiplier, i.e. the safety margin based on gravity and survival aerodynamic loads, that would cause local buckling is computed to be 1.33. A few characteristic samples of the results obtained from the nal discretization scheme are displayed in Fig. 4. It is worth noting the bending nature of local deformation patterns due to the essential lack of a signicant torsional moment. In order to illustrate the accuracy and practical usefulness of the above nite element buckling analyses, a comparison is made with the simplied analytical procedures specied in DIN 18800 (Part 4) [5] for cylinders and cones. The theoretical buckling stresses, according to DIN 18800 [5], can be computed only for simple cyl-

t= 12mm

t= 14mm t= 16mm
Fig. 5.

3 2 1

3.070 2.070 0.000

Simplied nite element model for seismic analysis.

1022

N. Bazeos et al. / Engineering Structures 24 (2002) 10151025

Table 2 Eigenvalues of wind turbine tower Rened model (Hz) First mode Second mode Third mode Fourth mode 0.937 7.400 14.631 18.984 Simplied model (Hz) 0.964 7.213 19.614

the wind turbine tower two different models are employed: (i) The rst one is the same nite element model used for static and stability analysis. Although this rened model requires signicant computational effort it is used for comparison purposes since it is expected to yield the most accurate results. (ii) The second one is a simplied multi-degree of freedom oscillator with 18 concentrated masses located at characteristic elevations, as shown in Fig. 5, and 108 degrees-of-freedom. It is a 3-D beam model that approximately maps the mechanical properties, e.g. cross area, moment of inertia, tapering, etc., of the real tower. The general-purpose computer program NASTRAN [13] has been used in conjunction with this model. 3.4. Fixed-base analysis The dynamic characteristics of the tower structure are considered rst on the assumption of xed boundary conditions at its base. The rst four eigenvalues of the tower, as computed by the above models, are listed in Table 2, and the corresponding eigenmodes are shown in Fig. 6. The results obtained by the two models are coincident for all practical purposes. The third torsional mode of free vibration cannot be detected by the simplied model, since only bending and axial deformation modes are considered. However, the contribution of the

the loaddeection curve necessary for the assessment of the inuence of the P effect on the overall deformation characteristics of the tower [12]. The results of such analyses [6] shows that the loaddisplacement diagram for the top of the tower has essentially linear behavior even at levels well above the static load design values. 3.3. Seismic analysis Tall steel towers are usually designed considering the effect of wind loads as the only source of environmental dynamic disturbances. The effect of earthquakes as a possible source of damage or loss of serviceability is often neglected, even in high-risk seismic areas. However, neglecting earthquake effects should at least be justied by means of appropriate methods of analysis in conjunction with the recommendations of applicable building codes. For the computation of the dynamic characteristics of

Fig. 6.

The rst four eigenmodes of the steel tower.

N. Bazeos et al. / Engineering Structures 24 (2002) 10151025

1023

Fig. 7. Discrete model for seismic analysis including soilstructure interaction.

torsional mode on the overall seismic behavior of this structure is negligible. 3.5. Elastic subgrade analysis The effect of the SSI on the dynamic characteristics of the structure under investigation can be assessed by introducing a set of discrete springs and dashpots at the soilfoundation interface, see, for example, Mulliken and Karabalis [14]. In this case the previously discussed nite element models are modied by adding stiffening and damping elements at the foundationsoil interface and taking into consideration the mass of the foundation, as shown in Fig. 7. In addition, a portion of the soil mass, considered to move in phase with the foundation should be taken into account [14]. For the purposes of this study semi-rock soil conditions are assumed at the particular installation site with

Poisson ratio 0.3, specic weight 21 kN/m3 and shear modulus 520 MPa. Two foundation designs are proposed for the particular tower structure, one is a 10101.80 m3 concrete block and the other is a 12.512.51.80 m3 concrete block. Either one of these footings is assumed to be rigid. Based on the previous soil properties, foundation geometries and the related formulae proposed by Mulliken and Karabalis [14], the spring constants and added soil mass are computed as listed in Table 3. The inuence of SSI upon the rst three eigenfrequencies of the tower structure are presented in Table 4. It is worth noting the signicant inuence of SSI on the rst, and by far most important, mode of free vibration, in spite of the relatively stiff soil conditions considered in this example. However, there is an even more signicant inuence of SSI on the higher modes of free vibration. The seismic response of the tower structure to the design seismic motions described in the previous section of this work have been computed using: (a) the rened nite element model along with standard modal superposition procedures and (b) the simplied nite element model of Fig. 5 in a direct time domain analysis. These two computations produce essentially identical results, representative samples of which are shown in Fig. 8. The domination of the seismic response of the structure by the rst mode of vibration becomes obvious. The maximum values of several design variables are listed in Table 5. In spite the low damping ratio (x 0.5%) and the elastic design spectra used for the development of the seismic motions, the wind turbine tower under investiTable 4 Inuence of soilstructure interaction on the dynamic characteristics of the tower structure Eigenvalue Fixed base SSI with SSI with conditions (Hz) foundation foundation 1010 m2 (Hz) 12.512.5 m2 (Hz) 0.964 7.213 19.614 0.873 1.935 3.736 0.712 1.935 2.751

1 2 3

Table 3 Spring constants and added soil mass for the three most prominent modes of vibration Footing 1010 m2 Spring constant Horizontal Vertical Rocking 14,070103 kN/m 17,457103 kN/m 371,429103 kN/m/rad Added mass 177103 kg 593103 kg 11,852103 kg m2 Footing 12.512.5 m2 Spring constant 17,588103 kN/m 21,821103 kN/m 725,446103 kN/m/rad Added mass 347103 kg 1158103 kg 36,170103 kg m2

1024

N. Bazeos et al. / Engineering Structures 24 (2002) 10151025

gation presents low levels of stresses, smax 50MPa, which is by far lower than the maximum stress produced in static analysis for pseudo-aerodynamic loads under survival conditions. Furthermore, based on the above dynamic analyses, a third approximate model can be developed for hand calculations. This simplied model is a one degree-offreedom oscillator in the form of a cantilever with a concentrated mass at its free end. The concentrated mass is equal to the mass of the tower and the mass of the nacelle, generator, gearbox, etc. at the top of the structure. The stiffness of this oscillator is computed from the total concentrated mass and the period of the rst eigenmode as it is computed by the rened model. The seismic response of the tower can be computed easily using this model in conjunction with either the response spectrum analysis, for the computation of the maximum response of the system, or any of the standard time stepping algorithms and the accelerograms described previously. The results produced by this model present a less than 15% deviation from the results obtained by the other two models used in this work.

4. Conclusions The design of the prototype wind turbine tower under investigation is proven to be successful as far as strength and operational requirements are concerned. The rened and simplied models developed for the static and seismic analyses are in remarkable agreement. For this type of structures the rened nite element models are necessary for the static and buckling analysis since a high level of accuracy is required at specic critical locations. Simplied analytical models, as those usually recommended by building codes, can predict with relative accuracy critical loads related to local buckling phenomena for this type of structures. However, if such models are used, great care should be exercised in incorporating appropriately the boundary conditions at the ends of each section of the structure. Approximate numerical models can also produce quite accurate results for seismic analysis as well. In most of these types of tower structures a seismic analysis produces no critical response.

Fig. 8. Seismic response along the three principal coordinate axes, versus time, at the top of the tower: (a) accelerations (mm/s2), and (b) displacements (mm).

Table 5 Maximum values of design variables at several elevations as produced by seismic analysis Node numberelevation (m) 1 (0.000) 2 (2.070) 4 (7.070) 6 (10.570) 8 (13.637) 10 (18.204) 12 (23.204) 14 (26.251) 18 (36.896) Max V (kN) Max M (kN m) Max sz (MPa)

132.884 129.634 126.971 120.325 113.178 106.473 102.736 98.983 85.892

3939.472 3671.001 3544.032 2787.272 2323.501 1841.506 1323.035 1011.394 0 (free end)

40.682 43.250 48.609 38.229 43.470 48.112 34.566 39.544 0 (free end)

References
[1] Germanischer Lloyd. IV-Non-marine technology, Part 1Wind energy. Hamburg; 1993. [2] Spera DA. Fatigue design of wind turbines. In: Spera DA, editor. Wind turbine technology. New York: ASME Press; 1994. p. 547 88 [chapter 12]. [3] Eggleston DM, Stoddard FS. Wind turbine engineering design. New York: Van Nostrand Reinhold, 1987. [4] Eurocode 3: design of steel structures-Part 1.4: general rules-sup-

N. Bazeos et al. / Engineering Structures 24 (2002) 10151025

1025

plementary rules for stainless steels, ENV 1993-1-4. Brussels; 1996. [5] German Standards (DIN 18800, Part 4). Berlin: Beuth Verlag GmbH; 1978. [6] Beskos DE, Karabalis DL, Hatzigeorgiou GD, Karamaneas H. Design of a wind turbine tower450 kW. Technical Report to the Center for Renewable Energy Sources, Department of Civil Engineering, University of Patras; 1996. [7] Karabalis DL, Beskos DE, Hondros ID. Design of a wind turbine tower and foundation450 kW. Technical Report to the Center for Renewable Energy Sources, Department of Civil Engineering, University of Patras; 1997. [8] Riziotis B, Voutsinas S. EPET II, Project #573: design of a wind turbine450 kW. Technical Report to the Center for Renewable Energy Sources, Department of Mechanical Engineering, National Technical University of Athens; 1996.

[9] Greek Seismic Code. Organization for Seismic Design and Protection. Athens; 1999. [10] Karabalis DL, Cokkinides GJ, Rizos DC, Mulliken JS. Simulation of earthquake ground motions by a deterministic approach. Adv Eng Softw 2000;31:32938. [11] NISA II. Numerically integrated elements for system analysis, Version 7.0. Troy, MI: Engineering Mechanics Research Corporation; 1997. [12] Chen WF, Lui EM. Structural stabilitytheory and implementation. New York: Elsevier, 1987. [13] Nastran, Version 67. Los Angeles: The MacNeal-Schwendler Corporation; 1992. [14] Mulliken JS, Karabalis DL. Discrete models for through-soil coupling of foundations and structures. Earthquake Eng Struct Dyn 1998;27:687710.

You might also like