Parameter Analysis Of High-Speedy Catenary Based On Neural Network Model

The purpose of catenary detection is to maintain its security and avoid the catenary-pantograph fault. Most faults' happening seems to be chanciness, in fact the hidden trouble range and its sequence can be found from catenary parameters before its operation. The parameters of catenary detection are respective independent and have inevitable relations. The occasion, which every parameter does not out of safe range and still have hidden truble frequently appears in practice. It is very important that integrates the relations among all parameters and estimates the security of catenary-pantograph system. The thesis researches on inherent relations among parameters and puts the fault forecast into practice.A method of establishing neural network model is choosed to analyze high-speedy catenary parameters. A predominance of neural network is that can take self-study through training. Neural network have the ability to conclude characteristics of all data through appropriate training by research the old detection data records. It is possible to find the rule that realated to running states by using neural network theory to dig mass detection data of catenary parameters. Expert system can be established to estimate and analyze the reasons of accidents in order to reduce faults' happening.The read, classification and store into different data files from the original data files should be operated before analysis. Data also can be filtered in this process. Studying on neural network theory and comparing with the respective predominance and application areas of several neural network models, linear neural network is choosed to analyze the relations among catenary parameters. Seven parameters are inputted into network training process in order to realize the system self-adaptation differentiation. Each input vector can adjust its weights automatically in training process. By comparing with the received network output and the goal output, finally possible fault point and its position can be got. A simple software of high-speedy catenary parameter analysis and fault forecast system is designed and based on those achievements above.