Multiobjective optimization of hot metal gas forming process to production of square parts using neural network and genetic algorithm

Hot metal gas forming is a process to form metals with low formability or high strength at room temperature such as aluminum, magnesium and titanium. With increasing temperature, the formability of these metals increases and the strength decreases. In this process, for producing parts with desirable properties, determination of optimal parameters is essential. In this study, hot metal gas forming process was simulated by using Abaqus software, and the influences of input parameters on the outputs were evaluated with simulation results. In order to validation of simulation results, the experimental test was carried out by using hot metal gas forming setup. For modeling hot metal gas forming process, an artificial neural network in Matlab software were trained by using data obtained from the numerical simulation. In this neural network, pressure, axial feeding and time were assumed as input parameters and the radius, minimum and maximum thickness were considered as output. In the next stage, this model was implemented as input function in multi-objective genetic optimization algorithm to obtain Pareto front and the optimum process parameters. Obtained optimum parameters include: pressure 13.07bar, axial feeding 0.78mm and time 65.73s and the values of corner radius, minimum and maximum thickness obtained from the optimum parameters are 5.49mm, 0.92mm and 1.57mm respectively.