B - spline wavelet packets and their application in the multiresolution non - stationary signal processing

Increasing requirements for the technical condition of machines induce the development of novel diagnostic methods for possible fault detection and identification in an early phase. Most of these methods are based on the processing of vibration signals. The classical methods often do not give full information about the actual condition of a machine. Therefore, it is necessary developing the appropriate diagnostic methods. Some of the promising signal processing methods are the group based on the Wavelet Transform (WT), which give a possibility for the effective diagnosing of non-stationary vibration signals in the time-scale domain. The generalisation of WT, the Wavelet Packet Transform (WPT), allows the extraction of additional useful diagnostic features from the signal. However, the effectiveness of diagnostics in the case of wavelet-based methods is determined by the selection of an appropriate wavelet function. In the present study, the author introduces new wavelet packets based on B-spline wavelets. A comparative analysis of their effectiveness was performed on non-stationary synthetic signals. The B-spline wavelet packets were applied for rolling bearing condition evaluation.