- 學位論文
應用模糊推論與類神經網路理論於離心式冰水機故障偵測與診斷策略之發展
Development of Fault Detection and Diagnosis Strategy Using Fuzzy Inference and Neural Network Theories for Centrifugal Liquid Chillers
摘要
本研究主要目的: 應用類神經網路及模糊推論之理論，發展離心式冰水機故障偵測與診斷策略(以下簡稱FDD策略)，並探討不同故障診斷策略對故障診斷率與誤診率之影響，以找出較適用之離心式冰水機故障偵測與診斷策略。研究結果發現：針對七種離心式冰水機漸變故障進行診斷，模糊推論FDD策略診斷率與誤診率分別為88.0%與12.0%；類神經網路FDD策略之診斷率與誤診率分別為93.7%與6.3%。與現有文獻陳[11]FDD策略(診斷率與誤診率分別為34.8%與65.2%)結果相較，提升了FDD策略診斷與降低誤診率。針對六種離心式冰水機漸變故障進行診斷，模糊推論FDD策略診斷率與誤診率分別為97.2%與2.8%；類神經網路FDD策略之診斷率與誤診率分別為98.5%與1.5%。與現有文獻凌[12]FDD策略 (診斷率與誤診率分別為36.4%與53.6%)結果相較，提升了FDD策略診斷與降低誤診率。模糊推論與類神經網路FDD策略均能提升故障診斷率以及降低誤診率，兩方法相較，類神經網路結果較佳，因此類神經網路為離心式冰水機故障診斷較適用策略。
並列摘要
Aim of this research is using theories of fuzzy inference and neural network to develop fault detection and diagnosis (FDD) strategy for centrifugal chiller. And explore to influence of fault diagnosis rate and misdiagnosis rate in different fault diagnosis strategy to find more suitable fault detection and diagnosis strategy for centrifugal chiller. For seven soft fault of centrifugal chiller, fault diagnosis rate and misdiagnosis rate of fuzzy inference FDD strategy were 88.0% and 12.0%; fault diagnosis rate and misdiagnosis rate of neural network FDD strategy were 93.7% and 6.3%. Compared with Chen [11] FDD strategy (34.8% and 65.2%), raised diagnosis rate and reduced misdiagnosis rate. For six soft fault of centrifugal chiller, fault diagnosis rate and misdiagnosis rate of fuzzy inference FDD strategy were 97.2% and 2.8%; fault diagnosis rate and misdiagnosis rate of neural network FDD strategy were 98.5% and 1.5%. Compare with Lin [12] FDD strategy (36.4% and 53.6%), raised diagnosis rate and reduced misdiagnosis rate.FDD strategy using fuzzy inference and neural network both could raised diagnosis rate and reduced misdiagnosis rate. Compared the two strategy, neural network is better. Therefore, fault detection and diagnosis strategy using neural network is most suitable strategy for centrifugal chiller.