Research On Trackside Acoustic Detection System

The wheel and axle of the train is key structure to train bogie. It is very importantto the safe operation of the train that the structure and function of the wheel and axle areintegrated. We should find the fault of wheel and axle as soon as possible. If we caneliminate the hidden danger in time, we can guarantee the customer safety preferably.The traditional detection method is detecting by skilled workers when the train isparking. It depends on the subjective judgment of the workers to determine the state ofthe wheel and axle, and it is difficult to find the early failure.This paper studies Trackside Acoustic Detection System deeply. The study tries toanalysis key links in the system and designs the concrete inspection solutions. Byresearching the sound caused by the running train wheel and axle, this article adopts themethod of feature extraction and pattern recognition to diagnose early fault in the trainwheel and axle.Acoustic signal is collected by the acoustic signal collecting system installed onboth side of the track, which complete acoustic signal sampling in high speed and highaccuracy and data transmitting in high bandwidth. Acoustic signal collecting systemmainly consists of PC104industrial computer, the analog signal collecting card andacoustic signal sensor. Acoustic signal collecting system sends the digital audio file tothe data processing server computer by the network.Acoustic signal preprocessing, feature extraction and fault classification is done bythe data processing server computer. Acoustic signal preprocessing part reduces thenoise in the acoustic signal by digital signal filter firstly. Then the acoustic signalpreprocessing part combining with the magnetic signals file identifies the location ofeach wheel and axle acoustic signal in audio file, and completes the separation andsynthesis of the each wheel and axle acoustic signal finally.Feature extraction and fault classification part is the focal point of this article, andthis is also the difficult point in the fault diagnosis system. Feature extraction part needto extract the characteristic signal of the wheel and axle from complex backgroundsignal adopting a reasonable algorithm. This paper studies many kinds oftime-frequency analysis algorithm, for example the Fourier transform, wavelettransform and Hilbert huang transform and so on. The rationality and validity of thealgorithm is verified by MATLAB simulation. Finally, the effective algorithm isprogrammed in C language.Fault classification part completes fault diagnosis using the method of patternrecognition. This paper adopts some artificial intelligence technology to complete thefault intelligent diagnosis including fuzzy recognition and neural network recognition. Artificial intelligence technology simulates the human way of thinking and doing things.According to the fault feature extracted from acoustic signal, fault classification part candiagnose the wheel and axle fault by computer.Trackside Acoustic Detection System is installed and debugged on the scene. Thissystem can complete the basic diagnostic function in the process of testing. As animportant part in the train safety monitoring system, TADS is able to intelligentlydiagnose early fault of the train wheel and axle online, which effectively improve thedeficiency of the traditional detection methods.