Fault Detection of a Water Loop Refrigerating System Using Multivariate Statistics

This paper describes the use of statistical fault detection techniques to improve energy management in a large water loop refrigerating system (WLRS). The water loop has one centrifugal chiller with specified capacity of 420TR in parallel with another screw chiller with specified capacity of 420 TR, feeding several buildings to control the temperature inside them through condenser units. Multivariate statistical fault detection techniques such as Principal Component Analysis (PCA) were used, in order to analyze the historical data from the system to detect abnormal situations, by means of statistical measures such as Square Prediction Error (SPE) and T2., Finally, using this expertise of plant engineers were used to determine the fault causes, and results will be used to prevent future abnormal conditions.