Calibration of Viscoplastic Model for Modified Polyphenylene Oxide Polymer Using Approximate Optimization Technique

In this study, mechanical properties of the modified polyphenylene oxide (MPPO) were measured using a uniaxial tensile test. Based on this, the modified two-layer viscoplastic model describing the mechanical behavior of MPPO was applied to conduct a parameter study concerning material constants through finite element analysis. Through this, meaningful material constants were determined for the material model, using which, a computational design of experiments using the interior central composite design (ICCD) was established. Then, the root mean square error (RMSE) for load and displacement between test and analysis was calculated. Subsequently, surrogate models for the relationship between material constants (input) and RMSE (output) were generated using a back-propagation neural network (BPN). The optimization problem of material constants was formulated and using the non-dominant sorting genetic algorithm-II (NSGA-II), an optimal solution of material constants was calculated. Based on these results, finite element analysis was performed to validate the model describing the mechanical behavior of the MPPO material by quantitatively comparing the results obtained using the conventional and modified two-layer viscoplastic model with experiment data.