March 29, 2014
By Hubert Lobo
The use of material data constitutes an important input to new-product development as companies use computer simulation to reduce real prototype testing. Simulating real life creates a need to approximate complex phenomena such as crash/drop, creep, fatigue and the non-linear behavior of metals, rubber, plastics, foams and composites. Failure to properly describe these behaviors in simulations results in poor correlation to reality and unacceptable risk, as simulation results are used in predictive design decision-making. In other words, there is a strong interest to manage this risk.
The problem is a complex one that requires the creation of the right kind of material data, as well as conversion of this data into the proper simulation inputs. The phenomenon being simulated must be understood; it is often a multi-variate situation where the material is simultaneously seeing a variety of effects such as temperature, rate and the environment, for example. With careful consideration, the variables can be reduced to a simpler set suitable for simulation.
For example, it would be fairly simple to realize that the effect of temperature on the properties of a sheet metal car door can be neglected in a crash simulation, but must be considered for the inner plastic door panel. Polymers used in biomedical implants would be best tested at 98.6°F and in a saline environment.
Material Testing and Calibration
Material testing must be performed to precisely quantify the physical behavior at the required conditions. The testing is often different from conventional testing, because the data are not used for comparative purposes but aim for absolute accuracy; the data represent the real material in the mathematical calculations of the simulation.
Accordingly, everything matters—including the validity of the actual experiment, the precision and traceability of the test instruments, and the expertise of the test technicians to ensure that the test is carried out correctly.
The material data must then be prepared for simulation. Most simulation programs do not simply accept material data as an input. Instead, they provide material models—mathematical formulas that represent different kinds of behavior, such as linear elasticity for metals; elasto-plasticity and viscoelasticity for plastics; hyperelasticity for rubber; rate-dependent, creep and fatigue models, to name a few. These models may be simple or complex, depending on what is being simulated.
In each case, the challenge lies in calibrating the model, a conversion process whereby the material data are converted into numbers that form the parameters of the model. Unfortunately, this step contains a high level of uncertainty, because of a lack of well-defined methodology and the complexity of the mathematical conversions.
While some CAE software solutions have provided tools to perform this conversion step, most rely on the skill of the CAE analyst to perform this task. A few material models exist that will actually absorb material data as input; for example the work of Paul A. Du Bois with respect to the Fu-Chang model for foams. Tools such as Matereality’s CAE Modelers can assist with this conversion process and write material cards, the esoteric data files that form the input for simulation.
Verification and Validation
Additional pitfalls for successful simulation include the setting of a number of parameters and variables—unrelated to the material data but often contained in the material card. Because of the large number of such uncertainties, many successful analysts will conduct a verification and validation (V&V) step to confirm that the material model is performing correctly.
A simulation is carried out of a simple experiment that explores the validity of the simulation. The simplest case is for the simulation to give back the raw experimental data of the test: for example, a stress-strain curve from the tensile test used to generate the material model.
More complex validations would seek to explore the validity of the model in alternate or multiple modes of deformation, like shear and biaxial tension. Such validations are usually reserved for more complex models that hold the capability of handling situations such as hyperelasticity for rubber and SAMP for plastic crash modeling. Validations quantify risk and give much-needed confidence to the analysts as they begin to use the material model to explore the real-life experimental space of the physical prototype that the simulation seeks to replace.
Hubert Lobo is president of DatapointLabs, LLC, Ithaca, NY. Send e-mail about this article to DE-Editorsmailto:[email protected].