Fault detection in air-conditioning systems using fuzzy models

Much of the previous work on fault detection in air-conditioning systems has used some form of quantitative model of the correct operation of the plant to estimate the states of the system so that they can be compared with measurements of its actual behaviour. In practice, it is often very difficult to obtain adequate representations of plant behaviour using quantitative models since, even if the underlying mathematical model is structurally correct, the values of many of its parameters are known only approximately. Qualitative methods have advantages in such cases. The use of fuzzy models can take account of the uncertainties associated with predicting the often highly nonlinear behaviour of faulty plants, can easily incorporate what expert knowledge is available about the symptoms of faults, and can generate a measure of the ambiguity associated with fault diagnosis