Calibration of Advanced Material Model for 3D Printing Materials using Optimization and Machine Learning Methods

Abstract

This thesis deals with some technique for optimization of cyclic plastic properties on specimens of 3D printed stainless steel 316L which has been manufactured by Selective Laser Melting (SLM) technology. The sample was horizontally oriented on printing base. The stress-strain behavior under uniaxial loading has been studied. For the ratcheting predictions the kinematic hardening rules of Chaboche and Armstrong-Frederick were used in conjunction with the non-linear isotropic hardening rule of Voce available in ANSYS. Another optimization technique was researched based on Neural Networks (NNs). The performance of two NN models have been analyzed in order to estimate the parameters in Chaboche model. The selected networks were: Echo State Network (ESN) and Extreme Learning Machine (ELM). Obtained results are presented and discussed including eventual improvement for the future.