Design And Implementation Of Autodetection System For Train Tail Devices

Safety is the precondition for an enterprise to run well, and the detectiontechnique is an ensurence for the devices to work normally and safely. With thetrains running speed highly promoted based on the advanced techniques and thesignificant breakthrough in the loading weight and information technology fields,most of the traditional organizational modes in transport and operational methodshave been changed greatly. Actually, for the cargo trains, discarding guarding car,canceling operational car's captital and using safety protection devices in the traintail(abbreviated to "train tail devices" in the follow) are all the importantreformations. However, because of the manual operation mode, which makes dataread and record randomly and trustlessly, and the diverseness of the detector'scapabilities, the current detecting ways for the train tail devices can't confirm theusing devices well worked. Making sure the reliability of the train tail devices andimproving the working safety have been an urgent problem to be resolved. Toaddress the challenge mentioned above, this thesis presents an auto-detection systemof the train tail devices. Based on the current detection devices, its means, and theintegration with the detection and management, every detection item in the systemwill be finished automatically and in sequence under the support from theintegration of the computers and mechanism-electronics. Fast detection speed,simple and standard operations, accurate and unique detection results, and the III四川大学硕士学位论文printability, documenting and statistics for each data are all the advantages of thissystem to help achieve scientific management. The core of this auto-detection system is the simulated working onlineenvironment for the train tail devices. The environment simulates the real workingconditions. According to the feedback signals of the detection, the system can judgethe status of the train tail devices in the environment. What's more, by usingdifferent means synthetically, the system can finally tell whether the train taildevices are right or out of work, and provide the possible reasons. The auto-detection system requires handling a lot of different feedback signals,such as voltage, current, pressure, feedback speech etc. Among those signals, thefeedback speech signal is a critical detection signal. This thesis discusses thebackground, working flow, system construction, technical solutions, detectionsignals and detection means of the auto-detection system of the train tail devices.Meanwhile, the thesis focuses on the application of the speech recognitiontechniques in the auto-detection system of the train tail devices. That includes thefeatures of the speech signals and the analysis,coding and preprocessing for thespeech signals. Long-term speech signals are non-stationary, however, the signalscan be regarded as relatively stationary signals in a short time. So firstly, accordingto the learning of the speech to be detected, we can get the speeches' features usingthe means of statistical recognition. Then we recognize those features by all kind ofcharacter detection methods including template matching, hidden Markov modelmethod and neural methods etc. And then we present the popular method in speechrecognition field: the hidden Markov model (HMM) method, and compare it withother relevant speech recognition techniques. Finally, we make a conclusion and aperspective of the auto-detection system of the train tail devices.