Prediction of Multi Responses in Radial Drilling Process using Mamdani Fuzzy Inference System

Abstract

Engineering problems often embody many characteristics of a multi-response optimization problem, and these responses are often conflicting in nature. To address this issue, the present work uses Grey- based Taguchi method to express surface roughness of drilled holes and drill flank wear into an equivalent single response grey relational grade. Experiments have been conducted in a radial drilling machine with five input parameters using L27 orthogonal array. It has been observed that combined response of flank wear and surface roughness is affected by almost all input parameters; however, drill diameter is statistically most significant one whereas Spindle speed is least significant input parameter. The prediction results were obtained via. Mamdani fuzzy logic model and BPNN and the corresponding results were compared. It is observed that Mamdani produces better result compared to BPNN in predicting the equivalent response grey relational grade. The advantage of mamdani fuzzy logic lies in the fact that it can take into account the uncertainty and impreciseness involved during experimentation. It is usually convenient for the practitioners to express model inputs in linguistic terms such as high, low, medium rather than expressing in quantifiable terms. The extraction of linguistic terms can largely reduce the chances of error, which is a constriction experienced in case of crisp values used in neural networks.