Normal Boundary Intersection Method Augmented Factor Analysis for Multivariate Process Optimization

摘要

Finding solutions to optimization problems with many response variables can be very challenging. Moreover, when response variables are correlated, not using multivariate techniques often lead to unsatisfactory solutions. This study applies factor analysis (FA) in order to reduce the dimensionality and eliminate the correlation between multiple responses. The extracted rotated factors are optimized through a normal boundary intersection (NBI) method, which is known as very effective in generating evenly distributed Pareto frontiers and finding optimal solutions even in non-convex regions. A case study of an AISI H13 hardened steel turning process using wiper tools is investigated. Three input parameters (cutting speed, feed rate and depth of cut) were considered. Eight responses were modeled through a central composite design (tool life, cutting time, mean roughness, total roughness, material removal rate, specific cutting energy, cutting force and cost of the process). The NBI method based on FA was successfully applied to the case study and achieved viable solutions.