Comparing statistical process control charts for fault detection in wastewater treatment

Area of Science:

• Process control and automation
• Statistical process monitoring
• Environmental engineering

Background:

• Effective fault detection is crucial for process supervision, particularly in critical applications like wastewater treatment.
• Statistical control charts, including Shewhart, cumulative sum (CUSUM), and exponentially weighted moving average (EWMA) charts, are standard univariate methods for fault detection.
• These methods exhibit varying performance based on fault characteristics.

Purpose of the Study:

• To evaluate the performance of Shewhart, CUSUM, and EWMA charts in detecting drift and bias sensor faults.
• To assess the impact of active process control on fault detectability.
• To determine the optimal method for sensor fault detection in wastewater treatment processes.

Main Methods:

• Comparative analysis of Shewhart, CUSUM, and EWMA control charts.
• Simulation of drift and bias sensor faults of varying magnitudes.
• Evaluation of fault detection using both controlled and manipulated variables within a process control system.

Main Results:

• The EWMA method demonstrated superior performance for both drift and bias faults, particularly excelling in detecting drift faults.
• EWMA charts achieved a low false alarm rate and reduced detection time compared to Shewhart and CUSUM charts.
• Monitoring manipulated variables resulted in lower missed detection rates than monitoring controlled variables, as set-point tracking can obscure faults.

Conclusions:

• EWMA charts are recommended for sensor fault detection in wastewater treatment due to their enhanced performance.
• Reducing fault detection time can significantly mitigate excess energy consumption associated with sensor faults.
• Monitoring manipulated variables offers advantages for fault detectability over controlled variables in actively controlled systems.