Preform optimization for reduction of forging force using a combination of neural network and genetic algorithm

Abstract

In this article, preform optimization in closed-die forging has been studied. The main objective of the optimization process is to reduce forging force by changing the preform shape as a variable parameter. In this respect, a combination of neural network and genetic algorithms has been employed. The finite volume method (FVM) is used as a simulation tool for the forging process. Simulation results have been used in a neural network analysis and a genetic algorithm to optimize the forging force. The neural network was used in several stages for modelling the system. The optimization process contains three stages. In stage 1 the preform shape is obtained by open die forging. Stage 2 is used for closed-die forging process to obtain the forging force. Stage 3 is for the filling of the die cavity. The geometrical parametric design process was used to accelerate the operation. An aeronautical forging component has been selected as a case study. The final results showed a negligible discrepancy of 0.3 per cent for forging force between neural network and direct simulation results. The optimization results prove that a reduction of forging force equal to 50 per cent can be achieved in comparison with an initial preform with 10 per cent extra volume.