Prediction of the mechanical properties of ceramic die material with artificial neural network and genetic algorithm

This paper presents a connectionist approach using back-propagation artificial neural network (BP) and genetic algorithm (GA) to predict the mechanical properties of ceramic die material. This method is using GA to optimize BP, including how to train the initial connection weights of network and to determine the threshold values. In the analysis, mechanical properties of ceramic die material are optimized while its composition is assumed to be predefined. Various output parameters are considered in the optimization analysis, including hardness, flexural strength and fracture toughness in the optimization analysis. And the input parameters are composition of a kind of material. Results from experiment simulation tool are utilized to train and test the BP-GA optimization approach. The analysis indicates that the BP-GA approach offers a robust and efficient method of optimizing the mechanical properties of ceramic die material when improved neural networks are utilized instead of neural networks.