Fault diagnosis for variable-air-volume systems using fuzzy neural networks

This paper presents a new method for fault diagnosis of variable air volume (VAV) air-conditioning systems. The method determines performance indices using self-organizing fuzzy neural networks (SOFNN). The SOFNN has two outstanding characteristics. Firstly, the learning speed is very fast and fuzzy rules can be generated quickly because no iterative learning is employed. Secondly, by using the pruning technology, significant nodes can be self-adaptive according to their contributions to the system performance. Consequently, the proposed method can achieve high performance with a parsimonious structure. Simulation results indicate that the SOFNN-based fault diagnosis method for VAV systems gives a very good performance in training speed and diagnosis speed and has high diagnosis rate.