Neural Network Multiobjective Optimization of Hot Forging

Hot Forging optimization depends on several factors, known with uncertainty: die pre-heating, geometry, tempering, workpiece temperature and shape, lubricant. There are also several objectives: quality, energy consumption and tool life. Global optimization methods require a numerous process evaluations to reach the optimum. While tests can be simulated by Finite Element Method (FEM), most of them were substituted by a Neural Network model. Neural Network training is less sensitive to problem dimension than standard Design of Experiments. The approach is assessed against the traditional Finite Element Optimization by exploiting a case study of a steel disc