A Data Driven Approach to Identify Optimal Thermal Parameters for Finite Element Analysis of Electric-Assisted Deformation Processes

Application of electric current pulses while deforming a material, commonly referred to as electric-assisted forming (EAF), is known to have desirable effects over its formability. In the finite element simulation of this electric-assisted deformation, the time-temperature profile is obtained by providing various temperature dependent thermo-physical properties of the material.Out of all the required properties for such analysis, effective heat transfer coefficient and Joule heat fraction are sensitive to the microstructure of the material, geometry of the specimen and the ambient conditions. Generally, these coefficients are identified by iterative FE simulations. A clear methodology to estimate these parameters has not been established yet. In the present work, a procedure is developed using a genetically evolved meta-model of the time-temperature profile, which is experimentally obtained from the pulsed current assisted uniaxial tension and compression tests. For this purpose, various multi-objective optimization techniques such as BioGP, EvoNN and cRVEA have been utilized to estimate the temperature profile in each case. It is shown that the tri-objective optimization procedure predicts the experimental temperature profile with greater accuracy (within +/- 5%) and is best suited to obtain the thermal modelling parameters of electric-assisted deformation, than other optimization techniques used in this work.