Research On The Methods Of Condition Monitoring Based On Vector Quantization For Train Brake Pad Temperature

At present, with the development of high-speed railway of China, it has become the preferred way for more and more people to travel. People rely on high-speed railway so much that higher request for operation safety has been put forward. And the train braking system, the most sensitive part of railway transportation equipments, is under the spotlight. The temperature data can reflect the working condition of train brake pad. Due to environmental factors and the lack of accurate analysis and judgment on the train operation data, the train remains potential safety hazard and low efficiency and furthermore difficulties in maintenance.Condition monitoring and fault diagnosing technology has made a lot of achievements at home and abroad after decades of research and application. It has helped avoid many serious accidents and brought economic benefits. As the breakthrough point of condition based maintenance, this thesis introduces the concept of condition monitoring and fault diagnosing technology. People can monitor and recognize potential failure in time by monitor and diagnosis train pad temperature. It provides the technical foundation for the efficient operation of train maintenance by using the condition monitoring and fault diagnose technology on train pad temperature.Meanwhile, for the similarity of speech recognition and pattern recognition, this thesis introduces Vector Quantization (VQ), an algorithm in speech signal processing field, into the management of train operation data. Thus it realizes condition monitoring on-line and fault diagnosing and prediction.According to actual operation situation, this thesis combines three steps in processing. Firstly data pretreatment step deals with the abnormal data; secondly data training step iterative trains different brake pad temperature by VQ algorithm and takes the sequence as a reference sample; Thirdly data recognition step compares brake pad temperature data with various reference samples and calculates Euclidean distance based on vector quantization programming to obtain diagnosis results and brake pad working condition.The results show that this method has a low false alarm rate and a quick time response. Moreover it also eliminates the influence of noise and jump and meets the requirements of train brake pad temperature monitoring in real-time and accuracy and also provides the basis for practical application and reference study value for other safety systems.