Lab test of three fault detection and diagnostic methods’ capability of diagnosing multiple simultaneous faults in chillers

Abstract

Chillers are large energy consumers which account for 20% to 40% of a facility's total energy consumption. The performance of chiller is often affected by faults introduced during initial installation or developed in routine operation. Over last two decades, much research has been performed on automated fault detection and diagnosis (AFDD) for the chiller systems. In the real world, a chiller is often affected by multiple faults. This research was for the first time to evaluate three promising FDD methods’ capacity of diagnosing multiple simultaneous faults in chillers. First, the three FDD methods were introduced, including the multiple linear regression (MLR) black-box model-based FDD method, the simple linear regression (SLR) model-based FDD method, and the decoupling-based (DB) FDD method. Second, multiple simultaneous faults test was conducted on a 90-t centrifugal chiller installed in a laboratory. Several common chiller faults were introduced into the test chiller. Third, the three FDD methods were tested to detect and diagnose multiple faults. And then, a detailed evaluation of the three FDD methods was performed. The test and evaluation results show that the decoupling-based (DB) FDD method has the best performance in dealing with multiple simultaneous faults.