Using Finite Element Process Simulation to Design and Simulate an Optimal Preform Mold with Neural Networks and Continuous Genetic Algorithm

Abstract

In the forging process, if the shape of the final part is complex, the raw material cannot be transformed into the shape of the final part in one forging step; therefore, the use of preform molds is essential. An optimal preform mold is a mold that can meet the design criteria. Design criteria include producing a defect-free part with the least volume of raw material, the least plastic strain, the least force required to perform the process, and also completely filling the final mold. In this research, using the ability of the continuous genetic algorithm to generate Cartesian paths, first several different preform mold shapes are produced using mathematical functions. Then, using finite element simulation of the process, the optimal mold selection criteria are calculated. From the information obtained from the simulation, the artificial neural network is trained so that it can predict the results of the simulation process. This network and design criteria are used for the objective function in the continuous genetic algorithm. Finally, the best shape of the preform mold is calculated by the continuous genetic algorithm, which is a mathematical function and is obtained by plotting this function in the Cartesian coordinates of the mold shape. Next, this method is used for a part with an H-shaped cross-section to examine its efficiency. The optimal preform mold for the part is calculated and its forging results are extracted by the continuous genetic algorithm. Also, finite element simulation is performed for the optimal mold to compare its results with the results obtained from the continuous genetic algorithm. Finally, the success of using this method was proven.