Perceptron network fault diagnosis on the shutdown of the fan in fan-coil unit

Abstract

Fault diagnosis is an important method of improving the safety and reliability of air conditioning systems. When the fan in fan-coil unit is shut down, there are temperature variations in the conditioned space. The heat exchanger efficiency is lower and the temperature in the room will change while the heat load of the room is stable. In this study, fault data are obtained in an experimental test rig. Thermal parameters as suction pressure and room temperature are selected and measured to establish a characteristic description to represent states of system malfunction. A new approach to fault diagnosis is presented by using real data from the test rig. Using the artificial neural network (ANN) in self-learning and pattern recognition modes, the fault is di- agnosed with the perceptron (one type of ANN model) suitable for pattern classification problems. The perceptron network is shown to distinguish types of system faults correctly, and to be an artificial neural network architecture especially well suited for fault diagnosis.