Ensuring an efficient turning process by means of desirability index optimization for correlated quality criteria

The desirability index (DI) is a method for multi-criteria optimization accepted widely in industrial quality management. The DI integrates expert knowledge into the optimization process by setting up desirability functions (DFs) of the quality criteria regarding their objective regions and aggregating them into a single performance index. However, the independence assumption of DFs rarely holds true in real turning applications, and a number of studies have been conducted proving the existence of dependencies between tool wear, surface roughness, tool life and cutting forces. As a consequence, the optimal solution obtained might be biased towards the group of performance measures, which have a high level of association (positive correlations). In this thesis, modfications of DI for handling correlated multi-criteria optimization are developed. By integrating principal component analysis (PCA) into the optimization procedure, the correlations of DFs can be eliminated, and the overall performance index, PCA-based DI, is formulated as a strictly monotonically increasing transformation of DFs; thus, the optimality of solutions can be guaranteed through the research of Legrand [26]. Apart from the PCA-based procedure, the weight-adjustment method provides an attractive alternative approach which is simpler and more exible, by introducing the weight-adjustment cofficients into the original formulas of DIs. The proposed procedures are demonstrated by means of case studies of a turning process optimization, and the optimization results are benchmarked with the traditional DIs. It has been shown in results that optimizations should be also subjected to the correlation information of performance measures. In addition, the procedure for determining correlation is found to be the second important key for a successful optimization