Multivariate Statistical Approaches for Advanced Process Monitoring and Fault Detection in Manufacturing

Authors: Tarun Parmar

DOI: https://doi.org/10.5281/zenodo.14507793

Short DOI: https://doi.org/g8v5j4

Country: USA

Full-text Research PDF File:   View   |   Download

Abstract: This study investigated the application of multivariate statistical techniques, specifically Principal Component Analysis (PCA) and Partial Least Squares (PLS), for advanced manufacturing quality control. Data from a large-scale electronic component manufacturing facility were analyzed using PCA and PLS models to evaluate their effectiveness in process monitoring and fault detection. The PCA model reduced the 50 process variables to eight principal components while retaining 85% of the data variance. It demonstrated 92% accuracy in fault detection with a 3% false positive rate. The PLS model pinpointed the crucial process variables that directly affect product quality, enabling accurate predictions of quality deviations. Both techniques significantly outperformed traditional univariate methods, reducing false alarms by 40% and improving the fault detection speed. The models exhibited robustness to moderate noise levels, suggesting their applicability to real manufacturing environments. While conducted at a single facility, which limits generalizability, this research provides evidence for the potential of PCA and PLS to enhance manufacturing quality control through improved process monitoring, rapid fault detection, and quality prediction. An integrated approach that leverages the strengths of both techniques could be a powerful tool for advanced quality management in complex manufacturing systems.

Keywords: Multivariate Statistical Techniques, Principal Component Analysis (PCA), Partial Least Square (PLS), Manufacturing Quality Control, Process Monitoring, Fault Detection

Paper Id: 231840

Published On: 2019-03-09

Published In: Volume 7, Issue 2, March-April 2019

Share this