Monitoring and fault detection in a reverse osmosis plant using principal component analysis

This paper presents a monitoring and fault detection system based on principal component analysis techniques (PCA) for a simulated reverse osmosis desalination plant. The classical fault detection techniques based on PCA are not suitable for the type of plants studied in this work due to the cleaning cycles necessary for correct plan running. These cleaning cycles periodically change the operating mode of the plant and these changes could be detected as faults. So a well known PCA approach designed for monitoring batch processes (U-PCA) has been adapted and applied in order to monitor this type of processes. The developed technique has been tested with several types of faults, obtaining good results. The ratio of false alarms has also been reduced in the nominal behaviour.