Integrated ANN-GA Approach For Predictive Modeling And Optimization Of Grinding Parameters With Surface Roughness As The Response

Grinding, the commonly used final finishing process across various industries for preparation of surfaces, uses an abrasive cutting wheel. Surface finish is the most commonly used index of final product quality in terms of aesthetics, corrosion resistance and others. The final surface finish of the grounded components depends on the cutting conditions of the grinding wheel and the machining parameters. Grinding wheel loading and wheel wear are the significant factors that determine the cutting conditions of the grinding wheel. Grinding wheel dressing is usually carried out to restore the original cutting conditions of the wheel. Speed, feed and depth of cut are the various machining parameters that affect the final surface finish. This paper aims to develop a predictive as well as optimization model by integrating Artificial Neural Network (ANN) with Genetic Algorithm (GA). Experiments were conducted on cylindrical grinding machine with Silicon Carbide grinding wheel. Speed, feed and depth of cut were selected as the three machining parameters with three different levels. Multilayer Normal Feed Forward ANN model of type 3-5-1 was considered for the prediction of surface roughness. Predicted values using the ANN model showed good agreement with the experimental values of surface roughness. Using ANN model alone could have the drawback that local minima based on initial parameters/training can be mistaken for the global optimum. Integrating Genetic Algorithm (GA) with ANN model overcomes this to a great extent. Such a hybrid technique can result in global optimal point of the machining parameters thereby leading to minimum surface roughness. The experimental results show the feasibility of the proposed method in the predictive modeling and optimization of grinding parameters. (C) 2017 Elsevier Ltd. All rights reserved. Selection and/or Peer-review under responsibility of International Conference on Materials Manufacturing and Modelling (ICMMM - 2017).