Machine Learning‐Based Bayesian Adaptive EWMA Control Chart for Industrial Process Monitoring

Conventional control charts that operate on set process parameters have a weakness in the fact that they fail to respond to monitor variations in the location parameter of process, and thus it takes a long time to detect or the detection may not be accurate. Conversely, adaptive control charts continuously change process parameters when online monitored by making them more responsive. In this research paper, a Support Vector Regression (SVR)‐based Bayesian Adaptive Exponentially Weighted Moving Average (AEWMA) control chart, as a novel method, is proposed to enhance the monitoring variations in the mean in industrial processes. This paper critically compares the performance of various kernel functions such as the linear kernel functions, radial basis function (RBF) kernel functions, and the polynomial kernel functions in the SVR framework. The analysis is undertaken to come up with a detailed comparison of the most effective kernel to be used in real‐time monitoring. In order to justify the suggested methodology, a practical example based on real world data in an industrial environment is provided and it proves the flexibility, accuracy and the increased sensitivity of the SVR‐based Bayesian AEWMA chart in effectively monitoring changes in process means. The findings reveal that the suggested solution is more effective compared to other conventional approaches, which makes it a valuable instrument in the current process management in flexible industrial settings.