You are on page 1of 16

Engineering Failure Analysis 8 (2001) 339±354

www.elsevier.com/locate/engfailanal

Numerical analysis of wheel cornering fatigue tests


U. Kocabicak *, M. Firat
University of Sakarya, Faculty of Engineering, Esentepe Kampusu, Adapazari, Turkey

Received 25 May 1999; accepted 2 July 2000

Abstract
In automotive engineering, the wheels are one of the most critical components and their function is of vital
importance n human safety. The cornering fatigue test is one of the traditional durability tests for wheel prototype
veri®cation. In this paper, a bi-axial load±notch strain approximation for proportional loading is proposed to estimate
the fatigue life of a passenger car wheel during the cornering fatigue test under plane stress conditions. The elasto-
plastic strain components are calculated analytically using the total deformation theory of plasticity. The input for the
load±notch strain analysis is the measured or calculated plastic strain state at the notch together with the materials
stabilised cyclic stress±strain curve evaluated with unnotched tension specimens. The damage accumulation is based on
the Palmgren±Miner rule. The methodology is implemented in a program called ``Metal Fatigue Prediction and Ana-
lysis'' (MFPA). The life prediction of a passenger car wheel during the cornering fatigue test is performed. The results
of the analysis is compared with two cornering tests on the same design. The result is very encouraging and the appli-
cation of the developed MFPA program provides time and the cost savings in the analysis of wheel cornering fatigue
tests. # 2001 Elsevier Science Ltd. All rights reserved.
Keywords: Fatigue assessment; Automotive design; Fatigue testing; Finite element analysis

1. Introduction

In automotive engineering, the wheels are one of the most critical components and their function is of
vital importance in human safety. With the changing requirements in the automotive industry in the last
decade, the wheel manufacturers have been taking increasing attention to lightweight designs by new
materials and manufacturing technologies in contradiction to durability concerns due to the complex
loading conditions on wheels. Knowing that an average design period is 6 months or more depending on
the requirements, iterations must be reduced at the physical prototype level and increased during virtual
prototyping with a reliable methodology to predict the durability parameters, such as the fatigue life,
during the sign-o€ tests.
In the fatigue evaluation of a wheel design, the commonly accepted procedure for wheel manufacturers is
to pass two durability tests, namely the radial fatigue test and the cornering fatigue test [1±3]. In addition to

* Corresponding author. Tel.: +90-264-346-0350; fax: +90-264-346-0351.


E-mail address: umit@esentepe.sau.edu.tr (U. Kocabicak).

1350-6307/01/$ - see front matter # 2001 Elsevier Science Ltd. All rights reserved.
PII: S1350-6307(00)00031-5
340 U. Kocabicak, M. Firat / Engineering Failure Analysis 8 (2001) 339±354

these tests, LBF-Germany [4,5] has designed the Biaxial Wheel Fatigue Test and its test machine is
intended to perform the combined radial and lateral cyclic loading to simulate the loads on road more
accurately.
The main objective in fatigue analysis is to predict the accumulated number of cycles at which a small
crack (such as 0.5 mm for mild steel) appears at a stress raising geometrical feature. The fatigue failure is
related to the initiation and subsequent growth of these small cracks under repeated loading and unloading
and the elasto-plastic stress state at the stress raising geometrical feature, or equivalently the notch, must be
analyzed in order to determine the deformation parameters describing the cyclic hysteresis loops [6±8].
Because of the lack of closed-form analytical solutions for the notch stress state in real life components, the
notch stress±strain approximation and related methods are used for the fatigue analysis.
The material deformation under cyclic loading is described with the Masing Plus Memory hypothesis
which is equivalent to the kinematic hardening theory of plasticity [9±11]. This model approximates accu-
rately the material stabilized deformation curve observed during cyclic testing with the unnotched speci-
men. It does not account for cyclic hardening, cyclic softening, cyclic creep or relaxation behavior of the
material during the initial few hundred cycles. The uniaxial load±notch strain approximation due to Neu-
ber and Seeger-Heuler is coupled with Masing Plus Memory hypothesis and uniaxial rain¯ow counting is
developed [12±16]. Due to restriction of the uniaxial stress±strain state assumption at the notch, the multi-
axial notch stress±strain approximations for proportional and non-proportional loading is developed [17±
27].
In this paper, a biaxial load±notch strain approximation for proportional loading is proposed to estimate
the fatigue life of thin sheet metals under plane stress conditions. The elasto-plastic strain components are
calculated analytically using the total deformation theory of plasticity [28,29]. The input for the load±notch
strain analysis is the measured or calculated plastic strain state at the notch together with the materials
stabilized cyclic stress±strain curve evaluated with unnotched tension specimens. The damage accumula-
tion is based on the Palmgren±Miner rule. Fatigue analysis software is developed and the fatigue failure of
a passenger car wheel under the cornering fatigue test is predicted. The results of the analysis are compared
with two cornering tests on the same design.

2. Critical stresses in a wheel

The stress distribution in a wheel depends on the mode of operation and the stress at any point in the
wheel is composed of prestresses and service stresses. The stress due to the manufacturing process, tire
pressure and assembly to the hub are the prestress on the wheel. The service stresses are result of the wheel
loads which are the vertical force Fv composed of static load due to the vehicle weight and dynamic forces
developed passing road irregularities such as potholes when driving straight, the lateral force FL due to the
quasi-static cornering forces as the most important ones. The longitudinal forces developed during braking
and acceleration as well as the wheel aligning moments are considered considered to be of secondary
importance. The highly stressed regions in a disk type wheel are the rim well, weld zone between disk and
cooling holes, bolt contact area and hat radius (see Fig. 1).

3. The cornering fatigue test

The cornering fatigue test is one of the traditional durability tests for prototype veri®cation. Fig. 2
represents the cornering fatigue test set-up. The aim of the test is to simulate the dynamic loading of the
wheel during cornering on the road. The relevant forces acting on the wheel are a combination of the ver-
tical load and lateral force developed between tire and ground. The two components of loading result in a
U. Kocabicak, M. Firat / Engineering Failure Analysis 8 (2001) 339±354 341

Fig. 1. Critically stressed regions on a wheel.

Fig. 2. Cornering fatigue test set-up.

rotating bending moment on the hub of the wheel. During the test, a constant or variable amplitude
bending moment is applied to the wheel. The bending moment is calculated as

Mb ˆ FR d ‡ FL R …1†

together with
342 U. Kocabicak, M. Firat / Engineering Failure Analysis 8 (2001) 339±354

FL ˆ FR …2†

where: FR= radial force acting on the wheel, FL = lateral force acting on the wheel,  = coecient of
friction between the ground and tire, R = radius of the statically loaded tire mounted on the wheel, d =
wheel o€set.
The acceptable service life of a wheel on a passenger car is considered as more than 10,000,000; however
the total time during cornering fatigue testing must be reasonably short and to achieve this the service
loads are increased with factors given in international standards such as SAE, DIN or ISO [1±5]. The
detailed procedure of the cornering fatigue test can be found in the standards.

4. Fatigue analysis of wheels

Disk type wheels are mechanical components produced by forming a thin sheet with stamping dies gen-
erally in three or four stages. In cornering fatigue test the whole component is under plane stress conditions
except at the bolt contact region and rim clamping region. Due to the wheel geometry, a reference section
for a nominal stress and even a pure notch cannot be de®ned. The only apparent stress raisers are the bolt
and the cooling holes on the wheel. Another point is the multi-axial stress state over the wheel which makes
the application of uniaxial local load notch strain approximations due to Neuber or Seeger-Heuler dicult
to apply. Consequently, multi-axial load±notch strain analysis must be employed. Regarding the loading in
the cornering fatigue test the only load acting is the bending moment through the hub of the wheel which
brings about a proportional loading.
In predicting the number of cycles in a cornering fatigue test, the crack initiation life is the primary issue
and the following statements should be addressed: (a) ®nding the crack initiation location(s) over the wheel
during the test; (b) evaluating the local stress±strain history at the critical locations.
In fact both of the above questions can be answered in an elasto-plastic ®nite element analysis using a linear
or non-linear kinematic hardening material law with isotropic von Mises yield criterion. Nevertheless, the ®nite
element analysis of the cornering fatigue test takes a few years in order to complete 106 cycles.
On the other hand, elasto-plastic ®nite element analysis with monotonic loading can be used to deter-
mine the realistic load±notch strain curve at the critical locations on the wheel. In the subsequent sections,
a stepwise procedure will be presented in which the elastic±plastic stress±strain history at any critical
location of a wheel during cornering fatigue testing can be approximated from the materials stabilised
cyclic load±notch strain curve and the bending moment history.

5. Load±notch strain analysis

The geometry of the wheel and the boundary conditions applied during the cornering fatigue test allows
two simpli®cations in conducting a load±notch strain approximation, the plane stress together with the
single external load, the cornering bending moment applied to the hub of the wheel. There is no apparent
reference section in the wheel and therefore the reference stress, a pseudo stress e , is de®ned as the stress
tensor calculated by elastic ®nite element analysis. Since the only external load is the wheel bending
moment (M) the following equations hold [17].

e
x ˆ Lx M …3†

e
y ˆ Ly M …4†
U. Kocabicak, M. Firat / Engineering Failure Analysis 8 (2001) 339±354 343

e
xy ˆ Lxy M …5†

where Lx, Ly and Lz are proportionality constants. During straining the rate of change of the deviatoric
stress components remains constant due to the single external load which is the bending moment. This type
of loading in plasticity theory is called proportional loading. Therefore, the total deformation theory of
plasticity can be used in approximating the elasto-plastic notch analysis for the wheel. Another point is the
plane stress state over the wheel except at rim clamping and hub bolt areas during loading. In this case the
total deformation theory can be solved analytically [17].
Hencky proposed the following total stress±plastic strain relations by which the total plastic strain
components are related to the current stress [28].

3 Sij
"Pij ˆ  "Pq …6†
2 q

where q is the e€ective stress and "Pq is the e€ective total plastic strain. Sij is the deviatoric stress compo-
nents. By (3)±(5), an e€ective pseudo stress can be de®ned as

q
e
q ˆ M L2x ‡ L2y Lx Ly ‡ 3L2xy …7†

The equivalent plastic strains are computed using the equivalent load±total notch strain curve given
functionally as
 
"tq ˆ f e q ˆ f Lq M …8†

The function f in (8) may be evaluated by a static elasto-plastic ®nite element analysis or by strain
measurement during the static loading.
By using the stabilised cyclic stress±strain curve of the material, the real e€ective stress q is determined
(see Fig. 3).
 
"tq ˆ g…a †; q ˆ g 1
"tq …9†

Then e€ective total plastic strain at the notch and plastic strain components are

q
"pq ˆ "tq "tq ˆ "tq …10†
E

Lx 0:5Ly p
"px ˆ "q …11†
Lq

Ly 0:5Lx p
"py ˆ "q …12†
Lq
344 U. Kocabicak, M. Firat / Engineering Failure Analysis 8 (2001) 339±354

Fig. 3. Material stress±strain curve and load±notch strain curve.


1 Lx ‡ Ly p
"pz ˆ "q …13†
2 Lq


3 Lxy p
"Pxy ˆ  " …14†
2 Lq q

Also inverting (6), the notch deviatoric stress components are computed as:
p
2 "ij
Sij ˆ  p qP …15†
3 "q

Since the normal stress must be zero the hydrostatic stress  can be computed from

z ˆ 0 ˆ sz ‡  …16†

The stress components are:

Lx
x ˆ q …17†
Lq

Ly
y ˆ q …18†
Lq
U. Kocabicak, M. Firat / Engineering Failure Analysis 8 (2001) 339±354 345

Lxy
xy ˆ q …19†
Lq

The total strain at the notch becomes:

Lx vLy e Lx 0:5Ly p
"tx ˆ "q ‡ "q …20†
Lq Lq

Ly vLx e Ly 0:5Lx p
"ty ˆ "q ‡ "q …21†
Lq Lq

   
Lx ‡ Ly e Lx ‡ Ly p
"ty ˆ v "q "q …22†
Lq Lq

Lxy e 3 Lxy p
"txy ˆ …1 ‡ v† " ‡ " …23†
Lq q 2 q q

where "eq is the e€ective elastic strain


q
"pq ˆ …24†
E

By examining the total strain components we note the following points:

a. All rational terms are constant over the wheel. By performing a linear static ®nite element analysis
with unit bending moment, all the rational terms may be processed to evaluate their distributions on
the wheel. This postprocessing step also enables one to ®lter the critical locations from the whole of
the component.
b. The only independent variable entering to the equations is the e€ective stress q . This is more obvious
if the material stabilised cyclic stress±plastic strain curve is given with a function
 n
q ˆ K "pq :

Following the work of KoÈttgen et al. the evaluation of load notch±strain analysis is called a e  based
approach and consists of the following steps [18];

a. Determine the proportionality constants Lx, Ly, Lxy in a cornering fatigue test with linear static ®nite
element analysis for unit bending moment.
b. Determine the e q "tq curve for the critical locations on the wheel for monotonic elasto-plastic ®nite
element analysis with isotropic von Mises yield criterion.
c. Compute the history of the stress tensor e ij …t† from (3)±(5). In the case of constant amplitude loading
they are
346 U. Kocabicak, M. Firat / Engineering Failure Analysis 8 (2001) 339±354
e
x ˆ M:Lx sin !t …25†

e
y ˆ M:Ly sin !t …26†

e
xy ˆ M:Lxy sin !t …27†

d. Apply this history to evaluate the e€ective strain history. Also by using the material stabilised stress-
strain curve, the history of plastic strain history is calculated.
e. Compute the stress component history by (13)±(16).

Having determined all the stress and strain components the ``Masing Plus Memory'' model for cyclic
deformation behaviour is used to determine the hysteresis loop developed in the material. The stabilised
cyclic stress±strain curve for the wheel material is described by the Ramberg±Osgood equation [12]

a   1=n
"ta ˆ g…a † ˆ ‡ …28†
E K

where the subscript indicates the alternating quantities and K, n are cyclic hardening coecient and cyclic
hardening exponent respectively. For the hysteresis branches the same function g is used with the sub-
stitution
 

" ˆ 2g …29†
2

with

" ˆ "max "min …30†

 ˆ max min …31†

The accumulation of damage is calculated with the Palmgren±Miner rule [12]. The hysteresis parameter
" and the number of occurrences of this parameter ni is determined. Then the total accumulated damage
is calculated as
Xk
ni
Dˆ …32†
iˆ1
Ni

where Ni is the read-o€ for each " on the constant amplitude strain±life curve of the material. The
functional description of constant amplitude life curve is given:
f
"ta ˆ "ea ‡ "pa ˆ …2N†b ‡"f …2N†c …33†
E

where: "a=strain amplitude, "ea =elastic strain amplitude, "pa =plastic strain amplitude, f =fatigue strength
coecient, b=fatigue strength exponent, "f=fatigue ductility coecient, c=fatigue ductility exponent.
U. Kocabicak, M. Firat / Engineering Failure Analysis 8 (2001) 339±354 347

Fig. 4. The ¯ow chart of developed MFPA program.


348 U. Kocabicak, M. Firat / Engineering Failure Analysis 8 (2001) 339±354

6. The developed program (MFPA)

In order to predict the number of cycles leading to failure in the wheel, based on the methodology
mentioned above a computer program called MFPA (Metal Fatigue Prediction and Analysis) is pre-
pared using the C programming language is developed. The ¯ow chart of MFPA is given in Fig. 4.

7. Damage calculation of a passenger car wheel for cornering fatigue test

The life of a wheel made up of RQST37 steel sheet of thickness 3 mm is estimated using the proposed
methodology. The following material properties are used in the calculation: E, elastic modulus=214 GPa;
v, Poisson's ratio=0.28; y , yield stress in tension=185 MPa.
The stabilised cyclic stress±strain curve is

a   6:14
"ta ˆ g…a † ˆ ‡ …34†
214:109 1:02:109

The constant amplitude strain±life curve of the material is

"ta ˆ 4:29417:10 3
N 0:053
‡ 0:5338  N 0:659
…35†

As the ®rst step, a linear elastic ®nite element analysis with unit bending moment is performed. The
proportionality constants in (3)±(5) are determined (see Figs. 5 and 6). Then an elasto-plastic ®nite element

Fig. 5. The Lx distribution on the wheel.


U. Kocabicak, M. Firat / Engineering Failure Analysis 8 (2001) 339±354 349

Fig. 6. The Lq distribution on the wheel.

Fig. 7. The e€ective plastic strain distribution on the 22.5o section.


350 U. Kocabicak, M. Firat / Engineering Failure Analysis 8 (2001) 339±354

Fig. 8. The load±notch curve at points 1, 2, 3.

Fig. 9. The hysteresis loop at point 1 developed initially.


U. Kocabicak, M. Firat / Engineering Failure Analysis 8 (2001) 339±354 351

analysis with isotropic strain hardening is conducted for a wheel bending moment of 3070 N m. In the
calculation, the loading is done in 25 increments and the residual force criterion is used in terminating the
equilibrium iterations with an error level of 0.001 [30]. Postprocessing the e€ective plastic strains (Fig. 7),
the load±notch strain curve(s) over the wheel is evaluated. The resulting curves at points 1, 2, 3 are given in
Fig. 8.
Examining the load±notch strain curves given in Fig. 8, the maximum relative di€erence between these
three curves is less than 10%. Although it is possible to use a di€erent load±notch strain curve at every
critical point over the wheel, if one applies a single universal curve during the cyclic deformation modeling
with the Masing Plus model, the computations are considerably reduced.
Having determined all the relevant coecients in (3)±(33) and the load±notch strain curves, all the stress
components and elastic±plastic strain components may be calculated for each load cycle and the corre-
sponding hysteresis parameter "a stored as a damage parameter. Fig. 9 shows the hysteresis loop developed
at point 1 and Fig. 10 shows the principle stresses at the same point. Fig. 11 represents the measured
dynamic strain history at three critical locations. For each "a stored, the relevant Ni is read-o€ from the
curve given by (35) and the current damage is evaluated by (34). The cycle at which the damage function
given by (32) equals 1, gives failure.
The prototypes are tested with a cornering moment of 3070 N m. The life times are determined as
489,780 and 512,450 resulting in an average value of 501,115 cycles. The failure locations are determined as
the cooling hole radius (around point 2) and hat radius region (point 1).

Fig. 10. The principal stresses at point 1.


352 U. Kocabicak, M. Firat / Engineering Failure Analysis 8 (2001) 339±354

Both locations are detected even with the linear elastic FEA, however the strain amplitude based on this
solution is 0.005967 and the life predicted is underestimated by 46%. On the other hand, following the
proposed methodology the number of cycles leading to failure at point 1 is estimated as 435,634, resulting
in 11% error, and at point 2 is 454,768, resulting in 11% error. The calculated cumulative cycles for failure
is given in Fig. 12.

Fig. 11. The measured dynamic strain history at three critical locations.
U. Kocabicak, M. Firat / Engineering Failure Analysis 8 (2001) 339±354 353

Fig. 12. The cumulative cycle count for three critical location up to failure.

8. Conclusion

A bi-axial load±notch strain curve approximation is proposed based on single proportional loading
condition. Pseudo stress coecient histories are calculated via linear static FE analysis with unit load.
Elasto-plastic FE analysis with isotropic hardening gives the e q "q curve at the critical locations. Total
stress±strain relations combined with the material stabilised stress±strain curve enables the calculation of
true stress histories. The hysteresis branch for each cycle is mapped onto the aq "qa plane according to the
Masing Plus Memory deformation model. The damage accumulation is based on the Palmgren±Miner
rule.
The methodology is implemented in a program called MFPA. The life prediction of a passenger car
wheel during the cornering fatigue test is performed. The predicted life shows only an 11% error with
respect to the physical test results conducted on two prototypes. The result is very encouraging and it can
be stated that the MFPA program will be a powerful tool to predict the fatigue life of passenger car wheels
during the cornering fatigue test.

References

[1] ISO 3006. Road vehicles Ð passenger car wheels Ð fatigue testing methods, vol. 2. Switzerland, 1976.
[2] Richard C, Rice M. SAE fatigue design handbook. 2nd ed. Warrendale (PA): SAE Publication, 1988.
354 U. Kocabicak, M. Firat / Engineering Failure Analysis 8 (2001) 339±354

[3] Wright DH. Testing automotive materials and components. Warrendale (PA): SAE Publication, 1993.
[4] Grubisic V, Fischer G. Procedure for optimal light-weight design and durability testing of wheels. Int J of Vehicle Design
1984;5(6):659±71.
[5] Grubisic V, Fischer G. Automotive wheels, methods and procedure optimal design and testing. SAE-830135, 1983.
[6] Seeger T, Heuler P. Generalized application of Neuber`s rule. J of Testing and Evaluation 1980;8:199±204.
[7] Seeger T, Beste A. Zur Weiterentwicklung von Naeherungsformeln fuer die Berechnung von Kerbbeanspruchungen im elastish-
plastischen Bereich, vol. 2. VDI-Fortschrittsberichte. 2;1997;1±56.
[8] Ho€man M, Seeger T. A generalized method for estimating multi-axial elastic±plastic notch stresses and strains. J Eng Mater
Technol 1984;107:250±4.
[9] Chu C. Incremental multi-axial Neuber correction for fatigue analysis. SAE-950705, 1995.
[10] Heyes P, Dakin J, St. John C. The assessment and use of linear static FE stress analyses for durability calculations. SAE-951101,
1995.
[11] Barkey ME, Socie DF, Hsia KJ. A yield surface approach to the estimation of notch strains for proportional and non-pro-
portional cyclic loading. ASME, J of Eng Mater Technol, Trans 1994;116:173±80.
[12] Dressler K, Hack M. Fatigue life estimation based on rain¯ow counted data using the local strain approach. Eur J Mech, A/
Solids 1996;15(6):955±68.
[13] Dressler K, Hack M, Kruger W. Stochastic reconstruction of loading histories from a rain¯ow matrix. ZAMM A Angew Math
Mech 1997;77:217±26.
[14] Dressler K, Gruender B, Hack M, Koettgen VB. Extrapolation of rain¯ow matrices. SAE-960569, 1996.
[15] Grunder B, Dressler K, Speckert M, Shaefer U, Oelmann B. Optimierte Erstellung von Prufzenarien. VDI Berichte-1335, 1995.
p. 175±88.
[16] Dressler K, Gruender B, Kottle H. Moeglichkeit der Berechnung in der Betriebsfestigkeitsanalyse. VDI Berichte-1153, 1994.
p. 43±59.
[17] Koettgen VB, Barkey ME, Socie DF. Pseudo stress and pseudo strain based approaches to the multi-axial notch analyses.
Fatigue Fract Eng Mater Struct 1995;981-18:1008.
[18] Koettgen VB., Barkey ME. Structural stress±strain analysis of non-proportional loading suitable for FEM postprocessing. Proc.
of Fatigue Design, Helsinki, 1995. p. 169±87.
[19] Brune M, Koettgen VB, Mawick G, Neureiter W, Reissel M. Abshaetzung der Betriebsfestigkeit des Schwesklages eines 5er
BMW basierend of FEM Analysen. CAD-FEM User's Meeting, 1997.
[20] Sonsino CM. Multi-axial fatigue of welded joints under in-phase and out-of-phase local strains and stresses. Int J of Fatigue
1995;17(1):55±70.
[21] Sonsino C.M. Overview of the state of the art on multiaxial fatigue of welds, vol. 5. Int. Conf. on Biaxial/Multiaxial Fatigue and
Fracture, Poland, 1997.
[22] Rupp A. Beanspruchung und Beanspruchbarkeit von Punktschweissverbindungen unter Schwingbelastung, PhD thesis, Uni-
versity of Carola-Wilhelmina Ð Braunschweig, 1992.
[23] Rupp A, Wessling U. Bemessung von Fahrwerkskomponenten auf Basis abgeleiteter Lastzeitfunktionen mittels Numerischer
Simulation, vol. 125. DVM-Berich. 1998. p. 59±79.
[24] Rupp A. Fatigue life prediction of thin sheet structures Ð theoretical and experimental results. IBEC'97, Internationales Messe
Congresscentrum, Stuttgard, 1997.
[25] Fischer G, Grubisic V, Streicher M, Wypich P. Verfahren zum Lebensdaurer nachweiss von Nutzfahrzeugblattfedern unter
Multiaxialer Belastung. Bauteil'94, Berlin, 1994. p. 253±65.
[26] Mueller A. Belastungmechanik und Simulationstechnik im Betriebsfestigkeitsversuch. GESA'96 Symposium in Schliersee, 1996.
p. 237±48.
[27] Seeger T. Betriebsfestigkeits auf der Grundlage oertlicher Beanspruchungen, Seminar. TH, Darmstadt. 1993, p. 120±36.
[28] Mendelson A. Plasticity: theory and applications. New York: MacMillan, 1968.
[29] Baeumel A, Seeger T. Materials data for cyclic loading. Amsterdam: Elsevier Science, 1990.
[30] ANSYS 5.5. Theory manual. ANSYS Inc., 1998.

You might also like