COMPARISON OF EXPONENTIALLY WEIGHTED MOVING AVERAGE CONTROL CHART WITH HOMOGENEOUSLY WEIGHTED MOVING AVERAGE CONTROL CHARTS AND ITS APPLICATION

The Exponentially Weighted Moving Average (EWMA) control chart is a widely used memory-type control chart known for detecting small shifts in process means. The recently developed Homogeneously Weighted Moving Average (HWMA) control chart modifies the weighting scheme of EWMA, giving more weight to the latest data and distributing smaller weights evenly to past data to further improve sensitivity. This paper compares the performance of EWMA and HWMA control charts on an iron pipe production process dataset. The methodology involves a two-phase analysis: Phase I for establishing in-control process limits (with normality testing, parameter estimation, and determination of optimal smoothing weights) and Phase II for monitoring new data using the established charts. The performance of each chart is evaluated using the Average Run Length (ARL) metric – specifically, the ability to quickly detect small shifts (ARL₁) while maintaining a low false alarm rate (ARL₀). The results indicate that the HWMA chart consistently achieves a smaller ARL₁ than the EWMA chart for small mean shifts without sacrificing in-control ARL, implying higher sensitivity to subtle process changes. Consequently, the HWMA control chart can detect small deviations in the iron pipe length more rapidly than the EWMA chart. These findings align with recent literature and demonstrate practical significance for quality control: the HWMA chart would enable earlier detection of process issues, allowing for quicker corrective actions in manufacturing. We conclude that the HWMA control chart outperforms the EWMA chart in this application, and we recommend its use for processes where small shifts in the mean are of critical concern. Additionally, we suggest further validation through Monte Carlo simulation and comparisons with other control chart methods (such as CUSUM or extended EWMA variants) to reinforce these conclusions for broader contexts.