COMPARATIVE PERFORMANCE ANALYSIS OF MEWMA AND DMEWMA CONTROL CHARTS FOR PROCESS MEAN SHIFT DETECTION USING MONTE CARLO SIMULATION

Abstract

This study compares the performance of the Modified Exponentially Weighted Moving Average (MEWMA) and Double MEWMA (DMEWMA) control charts in detecting process mean shifts. Effective clinker quality monitoring is important in cement manufacturing because process instability may affect product quality, operational safety, and environmental conditions. Parameter calibration was conducted using Monte Carlo simulation with 20,000 replications targeting ARL0?=370 (approximately 370 expected in-control observations before a false alarm occurs), where the optimal parameters were selected based on the smallest ARL1 and SDRL values. The evaluation used synthetic data under several outlier proportion scenarios and real C4AF clinker data consisting of 100 observations. The results show that DMEWMA generally provides better out-of-control detection capability and higher accuracy than MEWMA across various scenarios. In the synthetic data scenarios considered, DMEWMA generally produced higher monitoring accuracy and out-of-control detection rates than MEWMA while maintaining low false alarm rates. Furthermore, the application to clinker data shows that DMEWMA can detect more out-of-control observations without increasing the false alarm rate. These findings indicate that the additional smoothing stage in DMEWMA improves the preservation of process shift information, resulting in more reliable process monitoring performance.