Evolutionary Algorithm Approaches to Modeling of Flow Stress

Abstract

In order to reach a high quality of the metal forming processes and full functionality of the products, the properties of the material have to be determined as precisely as possible. In this article, the evolutionary algorithms are proposed for the determination of flow stress for steel X22CrNi17. Two evolutionary algorithm methods were used: genetic programming (GP) and genetic algorithms (GA). On the basis of experimental data obtained during torsion test, various different prediction models for the flow stress curve were developed independently by the GP and GA. To make a comparison, the models for flow stress were also developed by standard regression method. Accuracy of the best models was proved with additional measurements. The comparison between the experimental results, regression model results, and the solutions obtained by simulated evolution clearly shows that the GP and GA approaches are very strong evolutionary tools for solving similar problems.

Keywords:

• Evolutionary algorithms
• Flow stress
• Genetic algorithms
• Genetic programming
• Modeling
• Regression analysis
• Torsion test