ADAPTIVE CONSENSUS PRINCIPAL COMPONENT ANALYSIS FOR ON- LINE BATCH PROCESS MONITORING

Abstract. As the regulations of effluent quality are increasingly stringent and influent loads is, the on-line monitoring of wastewater treatment processes becomes very important to meet ever increasing effluent water quality. Multivariate statistical process control such as principal component analysis (PCA) has found wide application in process fault detection and diagnosis using existing measurement data. In this work we propose a consensus PCA algorithm for adaptive wastewater treatment process monitoring. The method overcomes the problem of changing operation conditions by updating the covariance structure recursively. The algorithm is also based on a consensus algorithm, which does not require any estimation for batch processes monitoring compared to typical multiway PCA models. With this method process disturbances are detected in real time and the responsible measurement are directly identified. The presented methodology is successfully applied to a pilot-scale sequencing batch reactor for wastewater treatment