Fan Fault Diagnosis Based On Multi-wavelet Theory

The fans are important equipment in industrial production. If the fan fails, serious consequences might be happened. So the fault diagnosis for fan has important significance.Wavelet analysis is a new theory. Compared with the Fourier transform, Wavelet analysis has a partial transformation of the time and frequency. So, wavelet analysis has a huge advantage in the digital signal processing known as "mathematical microscope". We all know compactly supported, which is the important nature of the signal processing. Single wavelet can not have these properties in the same time, so multi-wavelet is produced and multi-wavelet is widely used in fault diagnosis.At first this paper introduces the multi-wavelet theory, multi-resolution analysis and important properties, highlighting the advantages of multi-wavelet than single wavelet and Fourier transform. Second, because the multi-wavelet is multi-input output system and the fault signal is one-dimensional, pretreatment is essential before handling the fault signal. With the different pretreatment methods, the same multi-wavelet has different treatment effect. This paper studies SA4, GHM and CL multi-wavelet with different pretreatment methods for fault diagnosis. Simulation results show that the SA4multi-wavelet with the balance pretreatment has best noise reduction effect on fan failure signal. At the same time, we need set the appropriate threshold value to remove noise in the noise reduction process. This paper creates a new threshold value processing method in improving of the traditional soft and hard threshold method. The treatment of the new threshold is better by simulation,which comparing six common signal. Finally, collecting fan failure signal from some cement plant by PDES inspection equipment. Multi-wavelet decompose the fault signal and spectrogram is gat, we use spectrogram to analyse to confirm the point of failure. The conclusions of the analysis compare with automatic diagnostic results and then found that the diagnosis of wavelet theory used in this paper is more accurate. In addition, the same of analysis method is used to diagnosis fault signal of the main exhaust fan in some steel plant. We proved the effectiveness of the method once again.