Sensor fault diagnosis in a wastewater treatment process

Abstract

There are many sensors in a wastewater treatment process (WWTP) plant for monitoring process performance and condition. Sensor validation is essential to the success of process monitoring. In this paper, various sensor faults which can occur in WWTP are identified for taking proper remedial action at an early time. A proposed sensor fault isolation method is based on the variable reconstruction using principal component analysis (PCA). Even though several methods have been developed to identify sensor faults, they are only applicable to a static process. In other words, they cannot be successfully used to monitor severe dynamic processes such as WWTPs. We have removed this limitation by developing reconstruction methods based on a dynamic version of PCA. Artificial scenarios of sensor faults generated from the simulation benchmark have been used to validate the proposed sensor identifying methodology. Also, it is compared to a previous method to show its relative superiority in sensor fault validation in the WWTP.