A combined PCA and ANN approach for prediction of multiple responses in turning of AISI 1020 steel

Abstract

This dissertation presents an optimization approach for multiple responses (surface roughness, flank wear and cutting force) in turning of AISI 1020 steel with coated and
uncoated inserts using principal component analysis (PCA). Three controllable factors of the turning process were studied at three levels each viz. cutting velocity, feed and depth of cut. L27 Orthogonal array was used for conducting the experiments. Optimum parameters setting
to minimize surface roughness, flank wear and cutting force have been found out using Taguchi’s parameter design. Experimental results indicate that optimal factor settings for each response are different. Therefore, all the three responses are converted into a single response index through PCA approach. The process parameters are optimized with consideration of all the performance characteristics simultaneously. The analysis of variance (ANOVA) was used to find out the most influential turning parameters for multiple response problems. It is found that the cutting velocity has significant effect in producing lower responses
followed by feed and depth of cut. Since experimentation takes high amount of efforts, cost and time, it is prudent to propose a simple but valid model to predict the response. Therefore, an artificial neural network (ANN) approach, which works on experience of the modeler and
shop floor managers, has been proposed in this work. Neural networks offer a number of advantages, including requiring less formal statistical training, ability to implicitly detect complex nonlinear relationships between dependent and independent variables, ability to detect all possible interactions between predictor variables, and the availability of multiple training algorithms.