The Application Of Wavelet Support Vector Machine In Coal Mine Ventilator Fault Prediction

Mine ventilators has a very important role in the process of the coal mine production. It can not only guarantee the safe operation of coal mine, but also can effectively reduce the happening of coal mine gas accident. Once the fan fail, it may cause serious economic losses and even harm to personnel safety. Therefore mine ventilator fault prediction has a great significance to improve the production safety and ensuring the safety of personnel.The difficulty of the fault prediction was lies in the extraction of the characteristics. Fault information is often contained in mutation of a high frequency band signal. The traditional Fourier transform for mutation signal and non-stationary signal processing was flawed. But the wavelet analysis has a good effect in time domain and frequency domain. So first of all, it take the wavelet analysis instead of the Fourier analysis. It can be multi-scale decomposition of signal by the wavelet scaling and translation. The wavelet decomposition coefficients can be obtained in different frequency bands. Through the threshold method can remove the noise. Reconstruct the wavelet for each frequency band signal. Calculate the energy of high frequency wavelet decomposition coefficient by the wavelet coefficient of flat method. The signal energy was the Fault feature vector. The secondly, take the eigenvector input the support vector machine after the standardized treatment. To evaluate the reasonability inset dimension of support vector machine model by the FPE criterion. At the same time it use the Lagrangian multipliers and the kernel function into nonlinear problem instead of the complicated inner product operation. Then deduce the rational regression function. Training the input samples with the regression function. Then gradually get the forecasting model of support vector regression machine. It can undertake the fault prediction after the whole learning process when the training error can meet the requirements according to the parameter. Finally, take the Lab VIEW pc software to implement the forecasting model of the support vector machine. The construction of a basic program module was completed by constructing the data storge, wavelet denoising and support vector machine learning module. The nodes insert on the Lab VIEW software through the seamless connection between the Lab VIEW and the MATLAB can get the switch of the function. It can give the full play to realize the visualization and the interactive ability of the Lab VIEW. At the same time, the paper give the ventilator fault description with the Holospectra analysis method combining with the vibration data and the related experiment simulation.In this paper, the fan fault forecast put forward the corresponding theoretical analysis and some discuss. It has certain theoretical significance to improve the safe operation of mine fan and equipment maintenance. It also can applies to other types of large rotating machinery and equipment. It has a certain application value on its failure prediction.