Discount Off The Fault Detection Based On Neural Network Pattern Recognition Train

Under the train speed raising, the requirement for the safe is raising. The quality of the train brake system is the key of the train speed raising. The train coupling cock is one of the parts in train basic brake equipment, once it is closed wrongly, the air in the train pipe is blocked, the brake system can't work correctly, when the train is braked, the brake effect dropped, seriously it creates train accident, making huge loss for the country and people. Prevented train coupling cock mis-closing (TCCM) was listed as important study project by the train department in 1988.The paper introduced the train air brake system, analyzed its constitution and work theory. Based of this, draws out the place and effect of train coupling cock, then give out the problem and solve method of its work time, introduced the theory of detecting the TCCM. The paper analyzed the character and shortcoming of the detect TCCM equipment which was developed, pointed out the key technology of TCCM detecting. Because model the TCCM system is difficult, the paper try use neural network pattern recognition to detect TCCM. It possess cosmically collateral disposal ability, good fault tolerance ability, association of ideas ability and learning ability. We studied a key problem in a NN realizing PR system: pattern pretreatment and character pick-up, analyzed the K-L transformation which is the basic theory of realizing PCA.Based on which studied above, we gave out the core of the paper: the theory of realizing BP NN pattern classifier. Using the result of what we achieved, developed and designed a material system using NN to realize PR: the system of detect the TCCM. Recur to the MATLAB, using the ameliorated BP algorithm to train the swatch and finished the design of NN structure. The experimental research of the machine indicated it run steadily, detected the fault well and has ideal practicality. All these made clear the validity of the paper. In the end of the paper introduced the hardware and software design of our detect equipment.