Support Vector Regression Enhanced Adaptive EWMA Control Chart With Variable Sample Size

ABSTRACT This study introduces a novel adaptive exponentially weighted moving average (AEWMA) control chart that utilizes a piecewise linear function to dynamically adjust sample sizes based on the EWMA statistic. Traditional control charts, such as the classical EWMA, struggle with detecting small deviations and adapting to varying process shifts. The proposed method enhances the sensitivity of the control chart by incorporating support vector regression (SVR) modeling, which helps optimize the adjustment of sample sizes in response to process changes. Through Monte Carlo simulations, the performance of the proposed chart is compared with classical and variable sample size (VSS) EWMA charts, demonstrating effective performance in terms of run length and percentile values. The results highlight the chart's ability to detect both small and moderate shifts in process mean more rapidly and consistently. A real‐life application using cylinder bore diameter measurements further illustrates the practical advantages of the offered control chart, particularly in detecting subtle shifts often missed by traditional methods. This study validates the effectiveness of the suggested SVR‐based AEWMA control chart in precision manufacturing and dynamic quality control environments, emphasizing its potential to improve process monitoring and control.