Abstract

Digital Twin technology that predicts deformation, material change and other behavior occurring production phase is highly demanded since reduction of physical prototype with virtual prototype enables the cost reduction in addition to the shortening die design cycle. The key technology is the prediction accuracy of spring-back mode and amount. However this aspect of technology was not sufficient in case of high strength steel stamping cases in particular. The conventional elastoplastic constitutive equation assuming perfect elasticity inside of the yield surface has been used for elastoplastic analysis by nonlinear finite element method. On the other hand, the subloading surface elastoplastic constitutive equation has the basic structure that the plastic strain rate is always induced even in a low stress state. With the elasto-plastic traditional constitutive equation, it was difficult to precisely predict the spring back behavior after deep drawing metal forming process, in which material behavior during re-yielding in the reverse loading state is important. In particular, the amount of spring back of high tensile strength steels with initial yield stress of over 1 GPa, which are being used as automobile parts for weight reduction purpose in recent years, is bigger than conventional sheet metal and hard to predict. Data calibration technology is also important for the accuracy. We developed the system that calibrates material constant and other parameters for the Digital Twin with CADLM system by learning data and internally generates ROM system for the simulation input parameter calibration in order to minimize the deviation between simulation result and nominal value of physically measured shape. In this paper, we discuss on the implementation of subloading surface elasto-plasticity model into Marc implicit nonlinear finite element code and the effectiveness of sheet metal forming analysis using the subloading surface model in Marc which describes elasto-plastic material behavior even in reverse loading state accurately. This base Digital Twin model is subjected to optimization process for the accurate prediction performance enhancement.