A new and effective early fault diagnosis of rotor using order wavelet packet and markov chain model

The vibration signals at the start-up stage are non-stationary and non-Gaussian, and their diagnosis precision obtained with traditional diagnosis methods is not good. So we introduce the order wavelet packet and the Markov chain model that is based on particle swarm optimization into the early fault diagnosis of a rotor, thus proposing a new adaptive model of fault diagnosis. Sections 1 through 3 explain the early fault diagnosis mentioned in the title, which we believe is new and effective. Their core consists of; (1) we use the order tracking algorithm to carry out the resampling of the transient vibration signal, thus obtaining the diagnosis signal with equal angle distribution; (2) with the order wavelet packet, we decompose and reconstruct the equal angle distribution diagnosis signals and then extract their energy feature vectors at every frequency band; section 3 gives a five-step procedure for predicting the vectors with the Markov chain model. Section 4 conducts experiments on the early fault diagnosis of the rotor which uses vibration signals as its state signals; the experimental results, given in Tables 2 and 3, and their analysis show preliminarily that the short-term prediction results with our fault diagnosis model are very close to the actual values and have good prediction accuracy.