Studies Of The Rail Defect Detection Method Base On Vibration Signals

With the increase of the train operation speed, people are increasinglyconcerned about the safety of the train. Many railway traffic accidents are caused byrail defects, so the rail defect detection method are of great importance to the safetyof the train. Track inspection cars are the main rail defect detection tools at present,which detecting the rail every few months, and each time it will occupy the rail. Forhigh-speed railway, this detection method has some limitations. To solve thisproblem, a rail defect detection method base on acceleration sensors are proposed inthis paper. The acceleration sensors are distributed along the railway, from whichthe signals are sent to data processing center by wireless devices. Somecorresponding signal processing methods are used to attain real-time detection.An theoretical and experimental study of rail defect detection method waspresented in this paper.Theoretical research: Firstly, the vehicle-track vertical vibration system modelwas built by mathematic method. The corresponding vibration signals were acquiredby inputting the rail defects into the system. Secondly, after filtering preprocessingof the vibration signals, the corresponding feature parameters were acquired fromtime domain and time-frequency domain, with which the original rail defect featureparameter set was formed. Then the sequential backward selection method wasapplied to select the useful feature parameters. At last, the rail defects wereclassified and recognized by support vector machine.Experimental study: A wheel-rail experimental device was designed, which wasconsisted of wheel-rail contact system, speed control system and data acquisitionsystem. Some typical rail defects can be made on the simulation rail, and thecorresponding rail defect vibration signal was generated by wheel-rail contacting. Inorder to obtain more representative rail defect signals, the experimental device canset the simulation train speed, train weight and the measurement position. Sometypical rail defect signals were acquired by this experimental device. Then thesignals were processed by the method in this paper, and the rail defects are classifiedand recognized. The classification results of the experiments achieved theanticipated effects. The results show that the rail defect detection method in thispaper is very effective.