Machine learning based parameter-free adaptive EWMA control chart to monitor process dispersion

Abstract

Abstract Conventional control charts track changes in the process by using predefined process parameters. Conversely, during online monitoring, adaptive control charts modify the process parameters. To improve the process dispersion monitoring in various operational environments, this study presents an adaptive exponentially weighted moving average (AEWMA) control chart based on support vector regression (SVR). This study investigates the efficacy of different kernels such as linear, polynomial, and radial basis functions (RBF) within the SVR framework. By adapting the smoothing constant to the shift’s size in process dispersion, the suggested SVR-based AEWMA control chart makes better use of the strengths of the RBF kernel to identify shifts in the process dispersion. To demonstrate the method’s effectiveness, real-life data is used in a practical application, highlighting the adaptability and reliability of the SVR-based AEWMA control chart for monitoring process dispersion. The code and supplementary data set file may be found at ( https://github.com/muhammadwaqaskazmi/ARL-SDRL-Codes ).

Keywords

• Support vector regression
• Adaptive control charts
• Machine learning