Application of an Improved Analogical Reasoning Based on Similarity Measure Applied in Fault Detection and Diagnosis of Centrifugal Chillers

As there are some flaws in classic compositional rules of inference (CRI), analogical reasoning based on similarity degree (ARSM) was proposed. Most of ARSM models do not distinguish the difference between impreciseness derived by fuzziness of fuzzy concepts and uncertainty of observations and rules, which has adverse impact on the inference accuracy. On the base of ARSM computational models, an improved computational reasoning methodology is proposed. According to the impreciseness of antecedents and consequents, fuzzy production rules (FPRs) are classified into 7 different patterns. In terms of different patterns of rules, different reasoning models are employed to undergo the reasoning process. Then we applied the new-presented reasoning methodology in fault detection and diagnosis of centrifugal chillers, and respectively presented 2 examples of reasoning process with single rule and rule chain.