Computational Consistency of The Material Models
Computational Consistency of The Material Models
Abstract
The objective of this research was to investigate the effects of material models, element types, and boundary conditions
on the consistency of finite element analysis. Two cantilever beams were used; one made of stainless steel SUS301 3/4H
and the other made of polymer polyoxymethylene. The load–deflection curves of the two cantilever beams obtained by
experiments were compared to those obtained by finite element analysis, where the material models—including bilinear,
trilinear, and multi-linear—were used. Four element types—beam, plane stress, shell, and solid—were also employed
with the material models to obtain the simulated load–deflection curves of the cantilever beams. It was found that
bilinear material models had the stiffest behavior due to their overestimated yield strength. In addition, by applying a
finite displacement to simulate the grip of the cantilever beams, the discrepancy between the simulated permanent set
and the experimental set could be reduced from 80% to 5%. To sum up, both the selection of the material model and
the setup of the boundary conditions are critical for obtaining good agreement between the finite element analysis
results and the experimental data.
Keywords
Finite element analysis, material model, element type, boundary condition, consistency
Creative Commons CC BY: This article is distributed under the terms of the Creative Commons Attribution 4.0 License
(http://www.creativecommons.org/licenses/by/4.0/) which permits any use, reproduction and distribution of the work without
further permission provided the original work is attributed as specified on the SAGE and Open Access pages (https://us.sagepub.com/en-us/nam/
open-access-at-sage).
2 Advances in Mechanical Engineering
also contribute to the variation of the finite element that can affect consistency. Nevertheless, mesh density
solution.1 Commercial FEA software packages usually can be determined by observing the convergence of the
allow users to define the material model, adjust the solution with different mesh densities, so its effect was
mesh density, and choose the element type because they not considered here. The objective of this research was
are all related to the accuracy of the results.2 However, to find the effects of the material model, the element
how to choose the proper settings so that analysis type, and the setup of the boundary condition on the
results can be efficiently obtained, and with reasonable consistency of the FEA results with respect to the mea-
consistency, remains a challenge for many FEA users. sured results. Instead of using various constitutive
A previous study by Cifuentes and Kalbag3 demon- equations for the material models, bilinear, trilinear,
strated that the quadratic tetrahedral element might and multi-linear models were used. The conclusions
behave as well as the quadratic hexagonal elements drawn from this study can be used to improve consis-
regarding accuracy and CPU time. Li et al.4 also tency in future FEA.
reported that both tetrahedral element and hexahedral
elements were valid by comparing the simulation
results with the experimental data of a bogie frame. Materials and methods
However, it was found that the hybrid hexahedral ele- Material models
ment performed well, but the hybrid tetrahedral ele-
ment performed poorly for the simulation of the foot The materials used in this study were polyoxymethy-
and footwear under compression and shear loading.5 lene (POM), and stainless steel SUS301 3/4H. Tensile
Muccini et al.6 also showed that the linear tetrahedral tests were performed to obtain the stress–strain curves
element should be avoided during contact analysis due of these two materials, and then, the curves were used
to its accuracy problem. Regarding the mesh density, to generate their material models to for use in FEA. To
Langer et al.7 provided some guidelines for meshing: it perform the tensile tests, POM specimens and SUS301
was reported that at least 20 quadratic or 500 linear ele- 3/4H specimens were prepared according to ASTM
ments per standing wave were needed to achieve a solu- D638 (Type I) and ASTM A370 (sheet type), respec-
tion for eigenfrequencies with an error of less than 1%. tively, and three specimens of each material were tested
However, Shi and Liu8 studied different material using a universal testing system (Instron 3365, Instron,
models, such as those of Litonski–Batra, the power law, MA, USA). The stress–strain curves of POM and
Johnson–Cook, and Bodner–Partom, for simulating SUS301 3/4H were obtained as shown in Figures 1 and
orthogonal machining on an HY100 alloy. While some 2, respectively. Please note that the stresses and strains
material models could generate accurate predictions of used in this paper all refer to true stresses and true
cutting force and chip formation, they failed to predict strains.
the residual stress. From their study, it seemed difficult In FEA, three material models were used: bilinear,
to find one material model that could accurately predict trilinear, and multi-linear. These three models are com-
behavior in all aspects of interest. Lazarescu et al.9 also monly used when the material’s nonlinearity is consid-
used several constitutive models to generate yield sur- ered. The POM models are illustrated in Figure 1, and
faces for AA6016T4 aluminum alloy. Through hydrau-
lic bulge testing, they found that, to obtain good
agreement between the simulation and experimental
data, the constitutive model BBC2005 with all the input
parameters should be used. Thomas et al.10 applied a
cyclic loading using a punch on a simply supported cir-
cular plate to study various work hardening material
models, including isotropic, kinematic, mechanical sub-
layer, and Mroz. They found that the mechanical sub-
layer model was the most efficient, and had good
correlation with the experimental results. Amy et al.11
presented various error sources while performing FEA,
including material variability, dimensional tolerances,
and input parameter inaccuracy. They also reported
that the boundary condition was critical to the accuracy
of the FEA.
In this study, the measurement data were employed
as a reference to compare with the FEA results. As
mentioned previously, the material model, the mesh Figure 1. Experimental stress–strain curves and three material
density, and the element selection are critical factors models for POM.
Lee and Zhang 3
POM: polyoxymethylene.
POM: polyoxymethylene.
POM: polyoxymethylene.
Bilinear BEAM188
Trilinear PLANE183
Multi-linear SHELL181
SOLID186
Table 5. Deviations in the maximum deflection and the permanent set based on POM’s bilinear material model for various element
types (the values in the parentheses indicate the deviations in percentage with respect to the experimental data).
Maximum deflection (mm) –4.84 (–20%) –3.84 (–16%) –5.14 (–21%) –4.36 (–18%)
Permanent set (mm) –2.72 (–100%) –2.42 (–90%) –2.72 (–100%) –2.62 (–96%)
POM: polyoxymethylene.
Table 6. Deviations in the maximum deflection and the permanent set based on POM’s trilinear material model for various element
types (the values in parentheses indicate deviations in percentage with respect to the experimental data).
Maximum deflection (mm) 1.80 (7%) 2.46 (10%) 0.42 (2%) 1.29 (5%)
Permanent set (mm) 5.94 (218%) 5.55 (204%) 4.45 (163%) 4.64 (170%)
POM: polyoxymethylene.
Table 7. Deviations in the maximum deflection and the permanent set based on POM’s multi-linear material model for various
element types (the values in parentheses indicate deviations in percentage with respect to the experimental data).
Maximum deflection (mm) –2.54 (–10%) –1.54 (–6%) –3.44 (–14%) –2.54 (–10%)
Permanent set (mm) 0.18 (6%) 0.42 (15%) –0.52 (–21%) –0.22 (–9%)
POM: polyoxymethylene.
Table 8. Deviations in maximum deflection and the permanent set based on SUS301 3/4H’s bilinear material model for various
element types (the values in parentheses indicate deviations in percentage with respect to the experimental data).
Maximum deflection (mm) –1.02 (–8%) –0.75 (–6%) –1.40 (–11%) –1.40 (–11%)
Permanent set (mm) –2.86 (–100%) –2.72 (–95%) –2.71 (–95%) –2.86 (–100%)
Figure 5. Actual load–deflection curve of the POM cantilever Figure 6. Actual load–deflection curve of the SUS301 3/4H
beam and the simulated ones using various elements and a cantilever beam and the simulated ones using various elements
multi-linear material model. and a bilinear material model.
linear model is also between those of the bilinear and model for POM. Figure 6 shows the load–deflection
trilinear models, as shown in Figure 1. Table 7 lists the curves obtained using the experiment and finite element
deviations at the maximum deflection ranging from models, as based on the four element types. The devia-
26% to 214%, which are between the maximum tions in the maximum deflections calculated using FEA
deflections calculated using the bilinear and trilinear with the four element types ranged from 26% to
material models. The same situation occurred for the 211%, as indicated in Table 8. While the permanent
permanent set, which ranged from 221% to 15%. sets calculated by FEA were trivial, the actual perma-
Regarding the element type, all four element types nent set was 2.86 mm. This significant deviation was
behaved differently. mainly due to the overestimated yield strength of the
bilinear material model, as previously explained.
Regarding the element type, all four element types,
SUS301 3/4H BEAM188, PLANE183, SHELL181, and SOLID186,
This study extended this section to another material, behaved similarly.
SUS301 3/4H, which used the same approach as the Then, the trilinear model was used to calculate the
POM material. Regarding the SUS301 3/4H bilinear deflection and permanent set of the cantilever beam,
model, the results were similar to those of the bilinear and Figure 7 illustrates the results. The deviations in
Lee and Zhang 7
Table 9. Deviations in the maximum deflection and permanent set based on SUS301 3/4H’s trilinear material model for various
element types (the values in parentheses indicate deviations in percentage with respect to the experimental data).
Maximum deflection (mm) 0.86 (7%) 0.77 (6%) –0.38 (–3%) 0.04 (–0.4%)
Permanent set (mm) –0.21 (–7%) –0.54 (–19%) –1.32 (–46%) –1.02 (–36%)
Figure 7. Actual load–deflection curve of the SUS301 3/4H Figure 8. Actual load–deflection curve of the SUS301 3/4H
cantilever beam and the simulated ones using various elements cantilever beam and the simulated ones using various elements
and a trilinear material model. and a multi-linear material model.
the calculated maximum deflections, as listed in the permanent set between the calculated and the
Table 9, ranged from 20.4% to 7%, which were quite experimental results were large, which may indicate
consistent with the experimental maximum deflection. that there was an unknown factor that affected the
Next, the deviations in the permanent set calculation results of the permanent set.
ranged from 27% to 246%, as indicated in Table 9. One possible factor was the stress caused by the grip.
Compared to results of the POM case, as listed in In the cantilever beam experiment, a portion of the can-
Table 6, the calculated permanent sets based on the tri- tilever was firmly gripped to provide support. The stres-
linear model for SUS301 3/4H were relatively consis- ses of the specimens at the grip were generated, which
tent with the experimental results. Regarding the may have contributed to the large permanent set in the
element types, BEAM188 and PLANE183 behaved SUS301 3/4H’s case. Regarding simulation, we simply
similarly, as did SHELL181 and SOLID186. used boundary conditions with gripping portions per-
Finally, using the multi-linear material model for manently fixed at the initial position; thus, there were
SUS301 3/4H, the load–deflection curves were no stresses at the grip. To reduce the deviation in the
obtained, as shown in Figure 8. As the multi-linear boundary conditions between the experiment and the
model is stiffer than the trilinear model, as illustrated simulation, further simulations were performed using
in Figure 2, the load–deflection curves in Figure 8 the modified boundary condition, as discussed in the
shifted to the left, as compared to the curves in Figure following.
7. This trend was similar to that of the POM case. The
deviations at the maximum deflection, using the four Study on the setup of the boundary
element types, ranged from 21% to 28%, while the
deviations of the permanent set ranged from 262% to
condition
277%, as indicated in Table 10. Regarding the element In this section, a cantilever beam made of another
types, BEAM188 and PLANE183 behaved similarly, as material, SUS301 H, was analyzed to examine the setup
did SHELL181 and SOLID186. While the deviations of the boundary condition. The size of the beam was
in the maximum deflection between the calculated and the same as the one made of SUS301 3/4H, and the pro-
the experimental results seemed to be small, those of cedure was similar to that presented in the previous
8 Advances in Mechanical Engineering
Table 10. Deviations in the maximum deflection and the permanent set based on SUS301 3/4H’s multi-linear material model for
various element types (the values in parentheses indicate deviations in percentage with respect to the experimental data).
Maximum deflection (mm) –0.28 (–2%) –0.11 (–1%) –1.00 (–8%) –0.87 (–7%)
Permanent set (mm) –1.83 (–64%) –1.77 (–62%) –2.20 (–77%) –2.18 (–76%)
curve may be inappropriate. The yield strength of the 2. Diwan AG and Mahajan YS. Study of the effect of vari-
bilinear stress–strain curve may be much higher than ous parameters on the result of stress analysis obtained
that of the actual stress–strain curve, which may result using tetrahedral and hexahedral mesh elements. J Chin
in less deflection and permanent set. To use the trilinear Inst Eng 2017; 40: 101–109.
or the multi-linear stress–strain curve is a better choice 3. Cifuentes AO and Kalbag A. A performance study of
in adopting the material model. In addition, the devia- tetrahedral and hexahedral elements in 3-D finite element
structural analysis. Finite Elem Anal Des 1992; 12:
tion between the FEA results and the experimental
313–318.
results may be due to the over-simplified boundary 4. Li GJ, Wang WJ, An L, et al. Research on hexahedral
condition setup. Simply fixing the displacement of the and tetrahedral mesh applied to strength analysis of
boundary without considering the stresses induced by bogie frame’’, Appl Mech Mater 2014; 477–478: 150–154.
the grip may underestimate the permanent set gener- 5. Tadepalli SC, Erdemir A and Cavanagh PR. Compari-
ated at the boundary. son of hexahedral and tetrahedral elements in finite ele-
ment analysis of the foot and footwear. J Biomech 2011;
44: 2337–2343.
Declaration of conflicting interests
6. Muccini R, Baleani M and Viceconti M. Selection of the
The author(s) declared no potential conflicts of interest with best element type in the finite element analysis of hip
respect to the research, authorship, and/or publication of this prostheses. J Med Eng Technol 2000; 24: 145–148.
article. 7. Langer P, Maeder M, Guist C, et al. More than six ele-
ments per wavelength: the practical use of structural
Funding finite element models and their accuracy in comparison
with experimental results. J Comput Acoust 2017; 25:
The author(s) disclosed receipt of the following financial sup- 1750025.
port for the research, authorship, and/or publication of this 8. Shi J and Liu CR. The influence of material models on
article: The authors thank the Ministry of Science and finite element simulation of machining. J Manuf Sci E: T
Technology, Republic of China, for the financial support of ASME 2005; 126: 849–857.
this research under grant MOST 105-2221-E-011-055. 9. Lazarescu L, Ciobanu I, Nicodim IP, et al. Effect of the
mechanical parameters used as input data in the yield cri-
References teria on the accuracy of the finite element simulation of
sheet metal forming processes. Key Eng Mater 2013;
1. Langer P, Sepahvand K and Marburg S. Uncertainty
554–557: 204–209.
quantification in analytical and finite element beam mod-
10. Thomas TJ, Nair S and Garg VK. A numerical study of
els using experimental data. In: 9th international confer-
plasticity models and finite element types. Comput Struct
ence on structural dynamics (EURODYN 2014) (ed
1983; 16: 669–675.
A Cunha, P Ribeiro, E Caetano, et al.), 2014, pp.2753–
11. Amy RA, Aglietti GS and Richardson G. Accuracy of
2758. European Association for Structural Dynamics,
simplified printed circuit board finite element models.
https://paginas.fe.up.pt/;eurodyn2014/CD/papers/381_
Microelectron Reliab 2010; 50: 86–97.
MS14_ABS_1746.pdf