Adaptive EWMA Control Chart Using MLP‐ANN for Real‐Time Shift Detection

ABSTRACT Timely detection of process shifts in statistical quality control is crucial for maintaining production efficiency and quality. This study proposes a novel adaptive exponentially weighted moving average (AEWMA) control chart utilizing multi‐layer perceptron artificial neural networks (MLP‐ANN) to enhance shift detection capabilities. The proposed AEWMA‐ANN control chart is designed to monitor the process mean, serving as the primary location parameter by integrating an evolutionary machine learning approach. The effectiveness of this method is assessed in comparison with a convolutional non‐ANN parameter‐free AEWMA control chart and an AEWMA‐based support vector regression (AEWMA‐SVR) control chart. A key feature of the AEWMA‐ANN chart is its adaptive smoothing constant, which dynamically adjusts based on real‐time shift magnitudes detected by the MLP‐ANN model. The number of hidden layers in the neural network is optimized to minimize error, ensuring a robust and efficient detection mechanism. Performance evaluation is conducted through the Monte Carlo simulation method with 100,000 iterations, demonstrating the superior ability of the AEWMA‐ANN control chart. The findings underscore the potential of MLP‐ANN‐driven adaptive control charts for real‐time quality monitoring and process improvement.