Explainable multisensor fusion-based automatic reconciliation and imputation of faulty and missing data in membrane bioreactor plants for fouling alleviation and energy saving

Abstract

A membrane bioreactor (MBR) is a crucial wastewater treatment unit that requires continuous and precise monitoring to ensure stable operation and avoid energy loss. Sensor malfunction in MBR plants leads to missing and faulty measurements that negatively affect the control of membrane fouling and energy consumption. This study proposes an intelligent monitoring system to alleviate membrane fouling by the autonomous handling of sensor malfunctions in MBR plants using explainable AI (XAI) and a new multisensor fusion-based automated data reconciliation and imputation (MSF-ARI) approach. Missing and faulty data cases were used to validate the MSF-ARI approach in imputing the consecutive and point-to-point missing values and the detection, diagnosis, and reconciliation of the faulty data. Then, XAI analysis and an integrated biological-physical MBR model were utilized to evaluate the effect of MSF-ARI on membrane fouling and energy consumption. The results showed that the proposed MSF-ARI presents a superior missing data imputation performance (MAE = 0.31 mg/L) and a high reconciliation of the faulty measurements (MAE = 1.96 mg/L) along with a high detection rate (DRSPE = 100 %) and well-diagnosed fault groups. Applying MSF-ARI, the early membrane fouling was avoided by preventing the increased accumulation of the stable sludge cake, leading to a prolonged MBR operation by 10 days with an energy saving of 37.44 %. Therefore, managing faulty and missing data by MSF-ARI could contribute to the mitigation of membrane fouling and achieve a sustainable balance between operational time and energy consumption.