AE Signal De-noising Based On Improved Lifting Wavelet And Applied Research In Fault Diagnosis

Rotating machinery is the key equipment in the national production. It’smonitoring and diagnosis, safety and reliable operation, which plays an importantrole in raising the economic benefits of the enterprise and promoting thedevelopment of the national economy. The traditional monitoring and diagnosis isusually based on vibration signal, and the analysis method of the vibration signalsthat is difficult to obtain accurate fault characteristic information of equipment instrong noise background. The monitoring and diagnosis method that based on theacoustic emission (AE) signal has the advantages of high sensitivity, strongcapability of resisting noise and other advantages, and more easy to extractcharacteristic information characterization of running status of equipment. Itsapplication in the fault diagnosis of rotating machinery which can improve theaccuracy of fault diagnosis effectively.In the view of deficiency of traditional lifting wavelet, a new kind of improvedlifting wavelet transform method that suitable for rotating machinery AE signal wasput forward based on the analysis of the characteristics of rotating machinery AEsignal taking the rolling bearing as the representative.The method that rotatingmachinery AE signal denoising based on improved wavelet lifting wavelet wasstudied, and combining with the empirical mode decomposition in AE signal featureextraction and BP neural network in AE signal fault identification, the specific workis as follows:(1) The lifting wavelet improved algorithm for AE signal of rotating machinewas put forward.The improved lifting wavelet was proposed based on theoreticalanalysis and experimental research on characteristics of acoustic emission signal ofrotating machinery. It included in the decomposition process, the update operatorwas constructed adaptively according to the the local feature of AE signal withintroduction of local decision function; and using the genetic algorithm to optimizethe prediction operator.In the process of reconstruction, using adaptive thresholdde-noising on each layer of the AE signal which increase the lifting waveletdenoising performance. (2) The denoising method of rotating machinery AE signal based on theimproved lifting wavelet was studied.In the view of adaptive threshold denoisingmethod of parameter value p if too large or too small will affect the denoising effect.We can obtain the best values of the parameters in AE signal de-noising by studyingthe AE simulation signal. The de-noised results of the measured AE signal verifythat the improved lifting wavelet has a better denoising performance than wavelettransform and traditional lifting wavelet.(3) The characteristics extraction of rotating machinery fault AE signal wasstudied with improved lifting wavelet and empirical mode decomposition.Using theimproved lifting wavelet to de-noising AE signal, and then the correlation functionmethod is used to calculate the effective intrinsicthe mode functions after empiricalmode decomposition and envelope demodulation analysis, which can accuratelyextract the fault features of rotating machinery in the AE signal.(4) Combining improved lifting wavelet transforms and BP neural network toidentify the rolling bearing fault. The improved lifting wavelet as a former processorof the BP neural network which does the noise reduction pretreatment for AE signal,then extract the characteristic parameters of the AE signal after de-noising as theinputs of the BP neural network, reducing the training steps of the BP neuralnetwork and improving the recognition accuracy of the bearing fault.