Research On Magnetic Flux Leakage For High Speed Railtrack Detection

With the rapid development of railway, train safety is becoming more and more important. Therail health is the base of the train safety. Magnetic flux leakage (MFL), one of the ElectromagneticNDE methods, has simple structure, no pollution but high sensitivity. There is no need of contactwhile testing, so to make a rapid rail defect inspection possible. With these advantages, it is applied inthis paper to high-speed rail crack detection.Firstly, this paper compared different current rail defect detection methods. Then the principleof MFL at high-speed detection is introduced as well as typical signal characteristics. Due to themovement of detection device, the rail magnetization and eddy current are inevitable. To solve thefirst problem, Magnetic Barkhausen Noise (MBN) method is used to estimate the full magnetizationtime. And the effect of excitation voltage on the magnetization time is analyzed. In the promise ofMFL signals effective, magnetic yoke size and excitation voltage of the high speed MFL system arecalculated from the above in order to get higher detection speed. The excitation structure is optimized.Then the high-speed three-dimensional magnetic flux leakage testing platform is set up includingnon-destructive flaw detector, the reverse magnetic devices, three-dimensional sensors, signalconditioning circuitry and signal acquisition system. Some high speed MFL experiments aboutdifferent excitation structures are done on the platform. The experimental results are compared withthe theoretical value, and finally the yoke size and excitation voltage in the high speed MFL systemare determined. The highest speed which the system can achieve on above condition is determined.On the condition of full rail magnetization, Matlab software is used to analysis and process thehigh-speed MFL data. The defect parameters are studied through artificial neural network (ANN). Thespeed is one of samples of the BP ANN in order to eliminate the effect about the eddy current on thedata analysis and processing. And other defects are used to verify the neural network. The resultsshow that within a certain range of accuracy, the system is able to effectively detect and quantify raildefects. A health assessment of the rail is also achieved.Finally, the work having been done summarized, and some suggestions are given for furtherresearch.