Automated FDD of multiple-simultaneous faults (MSF) and the application to building chillers

Abstract

Heating, ventilation, air-conditioning, and refrigeration (HVAC&R) systems operated under faulty condition often result in extra energy consumption (up to 30% for commercial buildings) and cost, less comfort control and bad indoor/outdoor air quality, especially when multiple faults happening simultaneously. This study presents a novel hybrid strategy that combines support vector machine (SVM) and multi-label (ML) technique for the automated detection and diagnosis of multiple-simultaneous faults (MSF), and elaborates its application to a building chiller. One of the great advantages ML has against the mono-label (mL) technique is that no MSF data are needed for model training while a good FDD performance for MSF could be obtained. Two individual chiller faults and one of their combinations (an MSF) were investigated. Detailed studies on the use of three features sets and the training of the model with/without normal or/and MSF data were conducted and compared with the mL–SVM model. The results show that the ML–SVM model trained on the normal and two individual faults has an excellent performance, especially when the eight fault-indicative features (Feat8) were employed (correct rate over 99.9%). Feat8 behaves still excellent even when Gaussian white noise has been added to the test data.