Sensor selection in neuro-fuzzy modelling and fault diagnosis in HVAC system

Neuro-fuzzy modelling is a promising method to deal with large-scale and nonlinear dynamic systems. It can avoid the physical modelling and only deal with the measurements of inputs and outputs of the system. Generally, neuro-fuzzy modelling is composed of the selection of model structure, parameters and model validation. The quality of the model is largely dependent on the quality of the training data. How to select the input and output variables in the model is quite important. In this paper, not only the selection of sensor input in the model is discussed but faults can be diagnosed with the developed model. Sensor measurement plays a key role in system monitoring. Once a sensor fault occurs in the system, it should be detected immediately in case major catastrophe happens. A cooling coil subsystem in HVAC is used as a case study for input parameter selection in the model and fault diagnosis.