A UPCA-based monitoring and fault detection approach for reverse osmosis desalination plants

Abstract

This article studies and describes a monitoring, fault detection, and diagnosis technique based on the unfolded PCA (UPCA) approach and its application to a reverse osmosis desalination plant. The UPCA approach is normally applied to batch processes, but in this case, the UPCA approach is applied to a continuous process, which does not present a strict steady state. The classical principal component analysis (PCA) approach is not very suitable for this process due to the nonlinearities of this type of processes. The principal characteristics of PCA and UPCA methods are described. The different considerations and adaptations required to perform a UPCA monitoring tool applied to a continuous process, such as unfolding, alignment, and imputation, are also described and explained.