U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590
Federal Highway Administration Research and Technology
Coordinating, Developing, and Delivering Highway Transportation Innovations
|This report is an archived publication and may contain dated technical, contact, and link information|
Publication Number: FHWA-HRT-04-096
Date: August 2005
Evaluation of LS-DYNA Wood Material Model 143
PDF Version (8.16 MB)
PDF files can be viewed with the Acrobat® Reader®
1 - Developer's Introduction
The calculations and conclusions in chapters 1 through 8 of this evaluation report were conducted and documented by the developer of the wood material model, herein referred to as the developer. The calculations and conclusions in chapters 9 through 16 were conducted and documented by a potential end user of the wood material model, herein referred to as the user. Following these independent evaluation efforts is commentary, written by the developer in chapter 17, on the results of the user’s evaluation effort.
The wood model was primarily developed to simulate the deformation and failure of wooden guardrail posts impacted by vehicles. The primary features of the model are:
The behavior of the model is shown figure 1 for single-element LS-DYNA simulations that are conducted parallel to the grain. The simulations are linear to the peak in tension and shear, followed by post-peak softening. For these simulations, the softening is more brittle in tension than in shear. The simulation in compression is nonlinear because of the application of pre-peak hardening. No softening (perfect plasticity) is modeled in compression. A thorough discussion of the model theory is documented in the wood model manual.(2)
There are two methods of setting up the model input: The traditional method is to supply all material parameters (e.g., moduli, strengths, hardening, softening, and rate-effect parameters). A more convenient method is to request default parameters. The default parameters are obtained from laboratory data that are documented in the literature for southern yellow pine(5) and Douglas fir. The default parameters vary as a function of moisture content, temperature, and grade.
Figure 1. LS-DYNA simulations of southern yellow pine showing
brittle behavior in tension and shear, and ductile behavior in compression.
One limitation of the data available for setting default parameters is that the data are for clear wood (small specimens without defects such as knots), whereas real-world posts are graded wood (e.g., grades 3, 2, and 1, or DS-65). Clear wood is stronger than graded wood. Clear wood strengths cannot be used directly as input for graded wood. Our approach for overcoming this limitation is to apply strength-reduction factors to the clear wood data, which we call quality factors, to account for reductions in strength as a function of grade. This is a practical approach compared to the alternative approach of modeling each defect explicitly. Our methodology is to estimate the quality factors from correlations with the user’s static post and Forest Products Laboratory (FPL) timber compression data.
Other limitations of published laboratory test data for setting the default material properties include:
Our methodology is to estimate the missing material property values through LS-DYNA correlations with static post and bogie impact data provided by the user. Thus, the LS DYNA simulations discussed in this document not only serve to evaluate the material model, but also to set default material property values as well.
The evaluation of the wood model proceeded in two steps. The first step was the evaluation of the model as a user-defined material. This means that the model was hooked up to the LS-DYNA code as material model 42 via an interface. As the developer, we retained access to the wood model source code in order to enhance the formulation and adjust the default parameters during the evaluation process.
Once the evaluation was near completion and all of the default parameters were selected, the wood model was forwarded to Livermore Software Technology Corporation (LSTC) for permanent implementation into the LS-DYNA code. LSTC and the developer implemented the model into LS DYNA, beta version 970, as material model 143.
The second step was the evaluation of the wood model as material model 143 in the LS DYNA code. The objective was to check the permanent implementation to make sure that material model 143 produced the same results as the user-defined material. Adjustments in the LSTC implementation were made until agreement was achieved.
All evaluation calculations documented in this report were performed by the developer with the user-defined material model. Most were conducted using LS DYNA, version 960, on a DEC Alpha microprocessor using UNIX®. Subsequent calculations performed by the developer using material model 143 were conducted using LS-DYNA, version 970, on a personal computer (PC) using Microsoft® Windows®. Material model 143 calculations were in agreement with those performed by the developer via the interface.
Verification is a check on model implementation; it determines whether the material model has been implemented as the developer had intended (i.e., without coding errors). Stress-strain histories from single-element simulations were plotted to verify implementation of the wood material model.
Validation is a check on model theory; it determines whether the material model simulates real-world behavior. Multi-element simulations were compared to four sets of test data to initiate validation of the wood material model:
All of the test data discussed in this report were generated and documented by FPL and the user prior to performance of this contract. Comparisons of simulations with the user’s quasi-static and dynamic post tests are used to set the quality factors, fracture energies, rate effects, and frozen pine parameters used as default parameters in the wood model.
One might suggest that only pre-test predictions can be used to validate a material model. By this, we mean that the analyst is unaware of the measured results prior to the simulation. Accurate predictions (rather than correlations) certainly build the most confidence in a model. However, all calculations performed to date, and discussed in this report, were performed with the knowledge of the test results. This is because correlations with test results were used to set various default parameters. Future calculations performed by roadside safety analysts (such as the Centers of Excellence (COE) and the National Highway Traffic Safety Administration (NHTSA) National Crash Analysis Center (NCAC)) will assess the predictive capability and provide a more thorough evaluation and validation of the wood material model.