A Novel Approach for Integrated Fault Diagnosis Based on Wavelet Packet Transform

Integrated machine fault diagnosis is usually conducted by considering different types of signals so as to improve the accuracy of diagnosis. This paper presents a novel approach for integrated machine fault diagnosis based on the vibration signals alone. Wavelet packet transform is adopted to analyze the vibration signals, followed by the selection of best bases. We consider each best basis as a local site, then extract features from it and make a local decision using probabilistic neural networks. The local decisions from each best basis are fused to be a global conclusion using a weighted average method. The whole diagnosis process is implemented under a uniform framework. An experimental case shows that this approach improves the accuracy of diagnosis