Assessing the Application of Statistical Control Charts in Monitoring Process Stability and Performance Variation

Posted: Jul 17, 2019

• Matthew Wilson
• James Rivera
• Mateo Perez

Abstract

Statistical process control has long been recognized as a fundamental methodology for monitoring and improving manufacturing processes. The application of control charts, first introduced by Walter Shewhart in the 1920s, has evolved significantly over the decades, yet contemporary manufacturing environments present challenges that traditional approaches struggle to address. Modern industrial processes are characterized by increasing complexity, non-linear dynamics, and multi-scale variations that conventional control chart methodologies are ill-equipped to handle. This research addresses the critical gap between traditional statistical process control techniques and the demands of contemporary manufacturing systems by developing an integrated framework that combines wavelet analysis with multivariate control charting. The limitations of existing control chart methodologies become particularly apparent in environments where process variations occur across multiple time scales and where parameters exhibit complex interdependencies. Traditional univariate control charts fail to capture the multivariate nature of modern processes, while conventional multivariate approaches often lack the sensitivity to detect subtle, localized variations. Furthermore, the assumption of process stationarity underlying most control chart applications is frequently violated in real-world manufacturing scenarios, leading to increased false alarm rates and reduced detection capability. This research introduces a novel approach that addresses these limitations through the integration of wavelet-based multi-scale analysis with multivariate exponentially weighted moving average control charts.