Study On Fault Detection System For Axle Of Railway Vehicles

Railway vehicle is usually working in hazardous environment. The operational status ofaxle, which is the railway vehicle's main components to run, affects the safe operation ofthe railway vehicle directly. Over the years, the railway of our country has remained lowequipment-rate, high using-rate and high intensity-transport. In addition, because of theneed of the rapid development of economy, our country has raised the speed of the railwayvehicle, which leads to the railway vehicle's axle wearing and tearing seriously, the lifeshortened and accidents increasing significantly. So, it is very necessary and urgent todevelop the detection system because axle is related to the safety of the railway vehicle.This Ph.D. dissertation, taking wheel sets and axle of railway vehicle as the researchproject, is thought to be a theoretic and experimental study to apply detection technology,acoustic emission technology, electron technology, computer technology, signal processingtechnology and wavelet theory. Combining the science and technology developing project ofJilin Province: Intelligent Failure Detector of Railway Vehicle's Axle Online (serial number:20020326), it refers to the achievements the predecessors have attained and realizes todetect working performance of railway vehicle's axle and predict fatigue life of railwayvehicle's axle.1.The Railway Vehicle Axle Failure AE examination Acoustic Emission (AE) is the phenomenon that describes the transient elastic wavesgenerated by the rapid release of energy from localized sources within a material. It's aresult that the material and structure occurs deformation and fracture because of externalforce and internal force and then releases power in the form of elastic wave. The acousticemission technology is a nondestructive evaluation (NDE) material technology based onacoustic emission phenomenon to detect the degree of the internal fatigue.Although the NDE technology has been widely applied in the examination of fatiguematerial, there are some limitations. For instance, it fails to detect microcracks generatedinside of the materials, discriminate different fatigue accidents and reflect how the fatiguefailures affect the accumulated fatigue. Compared to other NDE technology, the acousticemission (AE) technology usually applied to detect active microscopic events such as crackinitiation, crack propagation, and record crack history in real time. The AE technology hasbecome the most important tool because of most materials' character of acoustic emission.The relation between the AE signal envelope frequency and characteristic frequency ofthe axle can distinguish the normal running and failure state of the axle. The variousinformation of the running axle can be reflected in the results of the different form ofenvelope frequency, and the relation between this different frequency AE signal andcharacteristic frequency of the axle is multiple frequency.The AE technology can detect microcracks and crack history, predict the damage degreeto forecast the accidents, so AE technology is an examination before failure. However theconventional vibration examination starts from the signal of failure that has been formed,so it is an examination after failure. In this opinion, the two different nondestructiveevaluation (NDE) material technologies play different role in the failure examination ofdevice.2. AE Signal Processing Based on Wavelet TransformThe aim of AE detection is to find the AE source and capture the information of it asmuch as possible. The technology of AE signal processing, which is how to get theinformation of the AE source from AE signal is not only the key point of the technology ofAE detection, but also the difficult point, because that AE signal is a kind of unpredictableburst instantaneous signal which includes high background noise. Therefore, the AE signalprocessing is substantively to capture the feature description of the AE source informationfrom the AE signal, which is half-baked, distorted, disturbed by vast noise.Signal processing play a very important role in AE detection. In the technology of AEnondestructive detection, from signal denoising to wave analysis, from AE sourceorientation to degree of damage analysis, each step can not go on without signal processingtechnology, therefore the research of signal processing technology is a hotspot in AEdetection. However, in the past actual researching, the effort about signal processing isfocused on the researching under the off-line condition, so it has no more demand on thealgorithm execution time and the realizable conditions of hardware. But in the process offailure on-line detection on the axle of railway vehicle, those questions change to beimportant. Firstly, if the algorithm of signal processing is complex, which canconsequentially result in vast calculated amount and long time of processing, the real-timeability cannot achieve the requirements of the system. Secondly, most of the hardware unitsof industrial device have equipped with embedded controllers, especially under theenvironments that the application conditions are hostile relatively such as railway vehicle.All the aspects of calculating word length, clock frequency, capability of floating-pointcalculating and programming tools of embedded controllers can not compare withcomputers, so the algorithm realized on embedded controllers can not achieve theanticipant effect, which can be realized on computers excellently. Therefore, the rapidityand effectiveness of AE signal processing algorithm is one key-point of AE on-linedetection technology.Wavelet transform with a capability of time-frequency analysis, can get a good effect ofdenoising. So wavelet transform is regarded as the tool of AE signal processing in thisthesis. For one thing, this dissertation makes an all-around analysis on the characteristics ofwavelet bases in normal use, then on these bases, a wavelet bases choosing rule and methodfor AE signal is presented according to the characteristics of AE signal in practicalapplication and the requirements of AE signal analysis. In the next place, based on thededuction of the function of frequency extension partition which is based on the Mallatalgorithm of wavelet multi-scale decomposition, this thesis indicates the relation amongmaximum decomposition scale, signal characteristic frequency of band and signal samplinglength, which makes the conception of more clearly, that wavelet analysis has acharacteristic which can make a frequency division analysis to the signal. In the end, on thedenoising effects, wavelet decomposition and reconstruction, wavelet transform thresholdand translation invariant are compared with each other through the research of experiments.At the same time, through anglicizing the calculated amount of three methods, a kind ofMallat algorithm based on discrete FFT is proposed. This algorithm can enhance thecalculated rate remarkably, and then achieve the demand of on-line filtering of AE signal.3. Axle Crack and Cumulative Fatigue ModelIn order to detect the axle failure of railway vehicle and predict residual axle life, theaxle crack and accumulated fatigue model must be set up. So the relationship betweenintensity factor of material and AE parameters should be established for using AEtechnology to judge the material crack and accumulated fatigue. In the past AE technologystudied on failure detection, AE counts are adopted to describe the equipment failure.Although the method of AE counts had widely been used in device fatigue detection andsome achievements have been acquired, it is limited by its inherent problem. AE countsdetection method only studies the counts of AE signals, the influence of the amplitude ofAE signals is ignored. An AE energy analysis is proposed to describe quantitatively crackfailure and accumulated fatigue of material in this paper. AE signals are decomposed inseveral frequency domains so as to denoise well based on wavelet, and then AE signalsenergy is calculated in its characteristic frequency, the quantitative relationships betweenAE energy and crack propagation and residual life of the material are established in the end.Experiment results show that the model of axle crack and accumulated fatigue based on AEenergy has proper structure and precision, which can meet the acquirement of the detectionsystem.The detection method of crack propagation and residual life of the material based onAE energy considers not only the amplitude of AE signal, but also the counts of AE signal.It is an improved method.4. Railway Vehicle Axle Failure Examination SystemTo make a comprehensive experiment on the failure of the axle, this dissertationconstructs a detection system, including the instrument for diagnosing the failure of theaxle of railway vehicle online and the analytical platform of failure based on virtualinstrument. The former completes the diagnosing of the failure online. By collecting AEsignals of the axle, the degree of the failure is accurately diagnosed, and then the alarminformation is given. Primary function of the latter is studying on positioning sources of thefailure and re-evaluation of the failure degree of the failure, using collection data from theformer or its own hardware.A hardware structure with DSP+FPGA is designed for diagnosing the failure of the axleof railway vehicle online. That structure connects with the certain features of the target,such as saving time, A/D transform control by FPGA, the AE signal with the identitycharacteristics of no-repetition and rapid decay and the bottom signal processing withoutDSP, by highlighting a 'ping-pang' FIFO for the real-time data collection module and aquasi-real-time data processing module with large capacity Ram butter designed forenhancing the function of the hardware as efficient as possible in the real-time system. Abrief on the choice of the sensor of AE and A/D converters operating frequency is given,the data for suspicious formation stored by USB disk, other sensors such as the speedsignals of the railway vehicle selected is discussed too.On the analytical platform of failure with virtual instrument, A hardware and thesoftware processing system is also designed based on the LabVIEW. PCI-2 data collectioncard of PAC company of USA is adopted to meet the requirements of AE signals collectionin hardware, By MATLAB script, the software platform can be achieved functions of AEsignals for time domain analysis, spectrum analysis, parameter analysis and waveletdenoising, etc. With the platform, this thesis studies on how to identify the state of axle andfailure location based on the collection of AE signals, and then detecting the typical AEsignals of axle in experiment. For the different state of axle, the result of experiment showsthat the envelope frequency characteristics of AE signals has obvious difference and hasrelations to character frequency. At the same time, the result of location experiment showsthat the time difference method based on wavelet transform and the first wave peak canenhance the precision of locating AE signals obviously.