Feature frequencies extraction and fusion of vibration noises of air-conditioner electromotor using adaptive resonance theory

There are three types of vibration noise in the air-conditioner electromotor. In order to get the reliable and accurate feature frequencies of each type of vibration noise, generally several spectra of same type of noise should be processed and the final feature frequencies are extracted and fused from them. Some principles of adaptive resonance theory being referred to, an algorithm is proposed to realize this object. In this algorithm several normalization computations, one threshold function and some positive feedback computations of spectrum vector form a close loop, simultaneously the fusion computation of different spectra is also embedded in this loop, in the course of loop computation, the feature frequencies can be strengthened while those interference frequencies can be greatly suppressed, ultimately a reliable and accurate feature spectrum can be formed automatically from many different spectra. The four spectra of each type of vibration noise of electromotor are processed and the results show that the feature frequencies of each type of vibration noise are extracted reliably and accurately by this algorithm.