A Hybrid Genetic Programming – Artificial Neural Network Approach For Modeling of Vibratory Finishing Process

Finishing processes are gaining importance over the last decades as manufacturers seek to improve process efficiency while meeting increasingly severe cost and product requirements. A number of researchers have employed conventional modeling techniques such as response surface methodology and linear programming but very few or none has paid attention to the non-conventional modeling approaches such as Artificial Neural Network (ANN), Genetic Programming (GP) and Fuzzy logic (FL) for studying the vibratory finishing process. Unlike conventional approaches, the independency of nonconventional modeling techniques on statistical assumptions ensures trustworthiness on the prediction ability of a model. The present study proposes a hybridized Genetic Programming-Artificial Neural Network (GP-ANN) approach for the formulation of mathematical models for the finishing process. The approach is based on error compensation which is achieved using an Artificial Neural Network (ANN) model in parallel with a Genetic Programming model. It is found that the hybridized GP-ANN models perform better than Genetic Programming models in terms of accuracy. The characteristics of the hybrid technique used are compared with response surface methodology. The results and analysis concludes that the GP-ANN approach is vital in those circumstances where data samples are fewer and can avoid the excessive GP runs for generating an optimal model. The ANN error model can also provide trustworthiness on the generalization ability of GP model whenever a new data sample is to be evaluated.