Model improvement with experimental design for identifying error sources in a computational model

Optimization-based model improvement has been introduced as a strategy to enhance the prediction ability of a computational model. It involves model calibration, model validation, and model refinement. Model calibration estimates the optimal value of unknown parameters. Model validation determines the validity of a computational model, and model refinement enhances a model to improve its accuracy. In model improvement, a variety of sources of errors in the observation and prediction can interrupt the model improvement process. The error sources degrade the parameter estimation accuracy in model calibration. When a computational model turns out to be invalid because of these error sources, model refinement is required. However, since model validation cannot distinguish between parameter estimation errors and modeling errors, the existing method is difficult to refine the computational model efficiently. Thus, this study aims to develop a model improvement process that identifies the leading cause of invalidity of a prediction. In this work, an experimental design method is integrated with optimization-based model improvement to minimize the effect of estimation errors in model calibration. Through use of the proposed method, after calibration, the computational model mainly includes the effect of unrecognized modeling errors. Two case studies are provided to confirm the efficacy of the proposed method: an analytical beam study and automotive wheel rim analysis.