Chiller Fault Diagnosis Based on Grey Similitude Relation Analysis

摘要

A new method of chiller fault diagnosis based on grey similitude relation analysis is proposed in this paper. Firstly, the comprehensive reference fault patterns are built by calculating weighted averages of reference fault patterns at multiple levels of severity. Secondly, the thresholds of grey similitude relation degrees are introduced to qualitatively identify) the belonging classes (special faults) of unidentified fault patterns based on similitude relation analysis. Thirdly, the results are quantitatively diagnosed to determine severity levels of identified faults. Genetic algorithm (GA) is introduced to optimize weighted indexes at multiple levels of severity and thresholds of grey similitude relation degrees. Seven typical chiller faults are discussed in this paper. This method is validated using the experimental data from the ASHRAE RP-1043. The results show that this method can be applied for chiller faults diagnosis effectively, and the accuracies of seven typical faults are 92.5-100%. Compared with the methods using one level of severity to build reference fault pattern and only relying on the principle of maximum relation degree, the proposed method significantly improves accuracies which are improved 65% at most (ExcsOil). And the proposed method decreases the iteration times of calculating relation degrees