Open Access Open Access  Restricted Access Subscription Access

Optimization of FE Crashworthiness Model for Sheet Molding Compound (SMC) with Extended Strength Distribution Model and Machine Learning



Sheet Molding Compound (SMC) possesses good mechanical properties and manufacturing flexibility. Nevertheless, SMC exhibits large scatter in mechanical properties and large difference in strengths under different stress distribution or loading conditions. As a result, the classical modeling techniques based on the mean tensile strength tend to significantly under-predict the flexural responses in finite element (FE) simulations. Probabilistic simulations with a unimodal tensile strength distribution had limited success in improving the predictions. In this work, the statistical distributions of the elastic modulus, tensile strength and compressive strength were considered with bimodal Weibull strength distributions. An optimization procedure for probabilistic FE models using artificial neural network (ANN)-based machine learning (ML) algorithm has been proposed. This procedure was examined through correlating simulations with experimental results for the 3-pt bending case. A set of optimum parameters was determined and subsequently examined in probabilistic simulations of tension, compression, 3-pt, and 4-pt bending cases. Simulations with this approach yielded low errors for all 4 cases and the results reached the steady-state within 20-30 iterations.


Full Text:



  • There are currently no refbacks.