Novel Fault Detection Approach of Biological Wastewater Treatment Plants

Abstract

It is well known that Exponentially Weighted Moving Average (EWMA) chart is designed to be optimal and efficient to quickly detect small faults. However, the classical EWMA can not perform well in the case of simultaneously large and small faults. To address this limitation, we propose to use an adaptive or a variable parameters control chart. Therefore, in this paper, we propose a novel approach, called particle filter (PF)-based adaptive EWMA (AEWMA) chart, with time-varying smoothing parameter lambda, to detect the fault in Wastewater Treatment Plant (WWTP) process. So that, the PF is applied to compute the residuals, and the AEWMA chart is used to detect the faults. The validation of the developed PF-based AEWMA technique is done using a simulated benchmark COST WWTP BSM1 model. The proposed PF-based AEWMA approach showed better detection abilities when compared to the classical EWMA and Shewhart charts.