Modeling and optimization for microstructural properties of Al/SiC nanocomposite by artificial neural network and genetic algorithm

Mechanical alloying process for synthesizing of Al/SiC nanocomposite powders was modeled by artificial neural network and then optimized by genetic algorithm. The feed-forward back propagation neural network model was used for predicting of the characteristics of the nanocomposite. These characteristics were the crystallite size, and the lattice strain of Al matrix. The aim of the optimization was to specify the maximum lattice strain and the minimum crystallite size of aluminum matrix that could be acquired by adjusting the process variables. Process variables included milling time, milling speed, balls to powders weight ratio that they were given as the input of the neural network model. Both modeling and optimization achieved satisfactory performance, and the genetic algorithm system proved to be a powerful tool that can suitably optimize process parameters. A comparison was made with an already carried out work; the model showed 37.6% improvement in error percentage of the crystallite size and 18.7% improvement in error percentage of the lattice strain of aluminum matrix.