Research On Acoustic Emission Detection Of Cracks In Straightening Process Of Shaft Parts

Shaft parts are prone to bending deformation after heat treatment.In order to ensure its straightness within allowable range,straightening must be carried out to ensure the radial run-out of each part of the shaft.In the course of straightening,axle parts are affected by the size of straightening pressure and the number of times of pressing.The material will be deformed results to be cracked.Therefore,the state detection and fault diagnosis of shaft parts in straightening process has significant practical significance for reducing the hidden danger of unqualified parts assembled on mechanical parts.At present,the crack detection of parts is mostly to detect the formed cracks and can only detect the external surface cracks,but it is difficult to detect the internal surface cracks of parts and the cracks generated during the straightening process.Acoustic emission(AE)detection is a passive method for crack detection,which is different from the traditional method.It only needs to receive signals from parts in the alignment process and analyze the signals to achieve the purpose of crack detection.Therefore,the main content of this paper is divided into three parts:1.The build of crack detection test-bed and signal acquisition,including senor selection and data acquisition device selection.The collected signal is de-noised by data acquisition unit in the straightening process of shaft parts.Wavelet transform has better de-noising effect on the noisy signal.The de-noising effect can be improved by setting appropriate thresholds.In this paper,the noise interference can be more effectively eliminated through the improved threshold function wavelet transform and retain crack signal.2.Empirical mode decomposition(EMD)method is used to decompose the signal and calculate the normalized energy of each intrinsic mode component(IMF),the total energy and energy entropy of each IMF.The normalized energy,total energy and energy entropy are taken as the characteristics of the signal to further diagnose the fault of the parts.3.The pattern recognition method of support vector machine(SVM)is applied to the fault diagnosis of shaft parts straightening process,which overcome the problem of need to obtain a large number of typical samples that traditional statistical pattern recognition method and artificial neural network,SVM can get better classification results on the premise of limited samples.With the normalized energy,total energy and energy entropy obtained from the feature extraction as the input of the vector machine,then,design a suitable classifier,the fault parts and qualified parts can be classified through the cracked and non-cracked parts as the output of the vector machine.In this paper,the method of wavelet transform,empirical mode decomposition and support vector pattern recognition is used to complete the crack detection in the straightening process of shaft parts through signal processing,feature extraction and state recognition.In the process of empirical mode decomposition,the concepts of total energy of signal and normalized energy of intrinsic mode component are introduced as the characteristics of signal,which improves the classification accuracy of support vector machine,reduced the rate of missing and false alarm of shaft parts in straightening process.