Abstract: There may be some kinds of hidden faults while ventilators are running during long time. It is very important to detect hidden faults of ventilators quickly for production and safety in mines. "Energy-faults" method is introduced in this paper, which is based on wavelet package decomposition and BP neural network. Character vectors which reflect different fault state of ventilators are extracted from different frequency segments with the technology of wavelet packet decomposition, and taking them input neural network as fault samples to establish the model of BP neural network. The fault states of ventilators can be identified by the BP neural network model. The results of research show that the faults diagnosis technology, based on wavelet packet and BP neural network, could exert both strongpoint, and it's an effective method of faults diagnosis by means of extracting mechanical faults characteristic. Meanwhile, it is an effective way to implement fault warning.