Comparison of Response Surface Methodology and Artificial Neural Network Technique for Prediction of Wear of 6061 Al-alloy – Sicp Composite

Metal matrix composites (MMCs) are among the very important engineering materials by which the combinations of various properties can be achieved. The MMCs have the great potential for use in automobile and aerospace industries for possessing high strength, stiffness and hardness. The reinforcement types used in the composites are continuous, discontinuous, short, whiskers and particles. The attention is, extensively, drawn in aluminum and its alloys as their reinforced with SiCp, in recent times, found increased demand in aerospace and automobile industries. This, in turn, necessitated the investigation of wear behaviors of these materials under different conditions as understanding tribological characteristics of MMCs leads to improved product design and enhanced durability. In this work, the wear behavior and the prediction models of SiCp reinforced composite were investigated under combined rolling-sliding conditions. The wear tests were carried out on MMC (10 vol. % SiCp) and its matrix6061 aluminum alloy. The effects of contact stress, time duration and surface speed on wear of MMC are investigated and the results are reported. One set of the obtained data is utilized for developing the two prediction model based on generalized regression neural network (GRNN) and response surface methodology (RSM). Another set is used for validation of these models. The choice of GRNN is motivated due to the requirement of small number of training data. It is observed in this investigation that GRNN not only able to model with minimum number of input data sets, but also performed better than the RSM based regression model.