Optimized tool design for expansion equal channel angular extrusion (Exp-ECAE) process using FE-based neural network and genetic algorithm

Abstract

Severe plastic deformation (SPD) methods have extensive capabilities to improve the mechanical properties of metals. Growing applications of SPD techniques in various industries demand comprehensive optimizations and parametric analyses. Expansion equal channel angular extrusion (Exp-ECAE) is recently proposed as an SPD process. In this research, effects of the geometrical parameters of Exp-ECAE tools on the distribution of effective strain on the product’s cross section are investigated. Then, the die geometry is optimized in order to achieve a product with desirable quality. In this regard, an artificial neural network (ANN), incorporated in a genetic algorithm (GA), is utilized. Required database, created by means of finite element (FE) analyses, is also used for ANN training. FE results are validated by experimental findings. The design of experiments is carried out based on nondimensionalized geometrical parameters. Finally, using two different approaches, the optimum die geometries are proposed for the Exp-ECAE process.