Livio Mazzarella

Temperature sensor signal reconstruction for failure detection of vapor compression system

Abstract Large effort has been made to fill up databases of failures and diagnosis for all types of technologies and they will remain open lists as systems are dynamically improving. This paper’s purpose is to provide support in decision making, quantification and diagnosis of failures in temperature sensor signal for a vapor compression system. The main challenge regarding this topic has been to distinguish and adapt methods to suit vapor compression systems. The temperature sensor signal failure was experimentally induced to a vapor compression system set-up, failure consequences were analyzed and three detection methods were evaluated: Principle Component Analysis, Fuzzy- Principle Component Analysis and Complex Fuzzy- Principle Component Analysis. All these methods are sensor reconstruction models, trained by the non-failure measured data to build the expected signal. Since faultless measured data are not always available, the possibility to use polynomial extrapolated data (as provided by manufacturer datasheets) is evaluated. The three selected methods showed to be suitable for similar heating, ventilation and air conditioning systems.