Fault detection and diagnosis of a refrigeration system using probabilistic neural network.

Aiming at finding an efficient way for the fault detection and diagnosis (FDD) of refrigeration system, the probabilistic neural network (PNN) is proposed to diagnose 7 types of typical faults for a refrigeration system. The establishment of the FDD model based on PNN and the processes of finding out the best spread value was elaborated in detail. The influence of sample size on the best spread value and the correct rate (CR) were explored. It was also demonstrated that the system-level faults were more difficult to be recognized by the model than the component-level faults. The comparison also has been done between the performance of the PNN and the prevailing back-propagation (BP) network. The results show that the overall diagnosis performance of the PNN is better than that of the BP network and the diagnosis of single training of the PNN is more reliable.