Navigating Process Drift: The Power of CUSUM in Monitoring Air Quality Processes and Maintenance Operations

Abstract

Nowadays, manufacturers face intense pressure to maintain a high standard of quality. Due to the damage to machine components, manufacturing processes degrade over time, resulting in substandard products. Generally, statistical process control tools such as control charts aid in identifying patterns and trends indicative of process changes. This investigation delves into the effectiveness of cumulative sum control charts using the sample mean and median as plotting statistics. Run-length measurements assess performance after the charts experience linear and quadratic drifts in non-normal setups under zero- and steady-state conditions. The findings reveal that Cumulative Sum (CUSUM) charts outperform zero-state monitoring compared to steady-state monitoring. Notably, the CUSUM chart for the mean is suitable for normal and Gamma distributions, exhibiting a greater ability for drift detection under biased and unbiased Average Run Lengths. This study offers valuable insights into enhancing manufacturing quality through effectively implementing and comparing Shewhart, Exponentially Weighted Moving Average, and CUSUM charts. By evaluating their performance under various conditions and comparing them with other control chart methods, this research provides valuable guidance for industries seeking to improve process monitoring and product quality. It is essential to acknowledge that the findings are based on specific experimental conditions and may not fully capture the complexity of real-world manufacturing environments. For practical purposes, the suggested charts are also applied to real-world case studies, including air quality (focusing on five metal oxide chemistry sensors: carbon monoxide concentration, non-metonic hydrocarbons, benzene, total nitrogen oxides, and nitrogen dioxide) and maintenance data (including air temperature, rotating speed, and equipment failure).