Enhancement of weak feature extraction in machinery fault diagnosis by using double density dual tree complex wavelet transform

In order to improve the deficiencies of classical discrete wavelet transform in extracting early and incipient machinery fault features, double density dual tree complex wavelet basis(DDCWT) is designed by introducing feasible redundancy in the filter bank. The filter bank of DDCWT consists of 2 scaling functions and four wavelet functions which form two Hilbert-coupled wavelet transform pairs, rendering DDCWT basis high regularity, nearly linear phase and nearly shift-invariance. In frequency partition, the central frequencies of DDCWT's subbands are embedded in transition bands of adjacent subbands of classical wavelet basis, thus achieving better performance in extracting features in transition bands of the latter. By applying DDCWT in a heavy horizontal lathe assembly examination, an assemble fault is detected. In addition, a de-noising method combing neighboring coefficients shrinkage strategy and stationary DDCWT is proposed and this method is applied in fault diagnosis of a gearbox in hot rolling mill. The de-noising method successfully extracts the two-site tooth fault features on the same gear.