We have present the viewpoint that uncertainties in predictions made by simulation codes should be the basis for code validation. A Taylor cylinder impact test is used to demonstrate the interpretation of uncertainty in terms of probability densitity functions. The first place to look for uncertainties in simulation code output are the uncertainties in the physics models that are incorporated in the simulation code. The Monte Carlo technique may be used to propagate uncertainties in the models into uncertainties in the code's predictions. Other sources of output uncertainties to consider include the numerical implementation of the physics models, especially the finite size of the finite elements, and aspects of the physics that are not accounted for. We point out that validation experiments may be used in an inference process to improve our understanding of the physics models, which is legitimately captured in terms of model-parameter uncertainties. This line of reasoning underscores the importance of conducting validation experiments that are thoughtfully designed to provide results that can be quantitatively compared to simulation codes.
Keywords: simulation code validation, simulation uncertainty, Monte Carlo technique, Taylor cylinder impact test, Johnson-Cook constitutive relations, model inference, uncertainty assessment, probabilistic network
Get full paper (pdf, 232 KB)
Return to publication list
Send e-mail to author at kmh@hansonhub.com