The wagons abnormal braking is a challenging problem for operators at marshaling yards. This paper develops an acoustic-based algorithm to detect the unreleased braking of wagons during uncoupling. Experiments have been done to collect the sound waves of wagons unreleased braking, as well as the sound waves of normal operation like train whistling and wheel vibration. The recognition algorithm is composed of fast Fourier transform (FFT), template matching and support vector machine (SVM) classification. Based on the sample data of different waves, the FFT is first performed to preprocess the original data for the frequency distribution from time-domain signals. Then, the major frequency characteristics of different sounds are extracted for SVM training, and each sound category corresponds to one acoustic template. Totally 19 SVM classifiers are generated through SVM training. Given a test data, at most 3 SVM classifiers will be activated according to the results of template matching. Finally, a case study of JSQ wagon and tank wagon has been performed to illustrate the proposed algorithm. Results show that the acoustic analysis and recognition can identify wagons unreleased braking effectively.

CICTP