Hybrid principal component analysis technique to machine-part grouping problem in cellular manufacturing system

Abstract

This article portrays a hybrid principal component analysis (PCA)-based technique to construct production cells in cellular manufacturing system (CMS). The key problem in CMS is to recognise the machine cells and corresponding part families and subsequently the formation of production cells. A novel approach is considered in this study to systematise a hybrid multivariate clustering technique based on covariance analysis to form the machine cells in CMS. The intended technique is demonstrated in three segments. Firstly, a similarity matrix is developed by exploiting the covariance analysis procedure. In the second stage, the PCA is utilised to identify the potential clusters in CMS with the assistance of eigenvalue and eigenvector computation. In the last stage, an adjustment heuristic is adopted to improve the solution quality and consequently the clustering efficiency. This article states that, the addition of the adjustment heuristic approach into a traditional multivariate PCA-based clustering technique not only enhances the solution quality significantly, but also downgrades the inconsistency of the solutions achieved. The hybrid technique is tested on 24 test datasets available in published articles and it is shown to outperform other published methodologies by enhancing the solution quality on the test problems.