Research On The Current Curve Of S700K Switch Machine Based On FrFT And Improved BP Neural Network

Currently, the construct of high-speed railway in our country has the biggest scale of high-speed railway system around the world.High-speed railway makes railway transportation more efficient. At the same time, it puts forward higher requirement for safety and reliability of related railway equipments.S700 K three-phase alternating dynamoelectric switch machine is a new switch machine which is introduced from German, the Siemens Company after speeding-up. As a central device along the route, the working condition of swtich machine is always a key monitoring object. The microcomputer monitoring system is used to monitor and assess working condition along the route in our country.The microcomputer monitoring system offers the information of working condition to the maintenance personnels, such as current curves and working time. Current curve includes starting current, indicating current action current and so on.However, the microcomputer monitoring system only offers the working current and time information of switch machines to our maintenance personnels without the function of intelligent analysis. The accuracy of maintenance personnels’ analysis on switch machines’ fault usually depends on personal working experience and the professional knowledge. The miscarriage of justice occurs sometimes. This method doesn’t meet maintenance personnels’ efficiency requirement of Chinese railway. Therefore, a new method is put forward that action current curve of the switch machine is choosen to be the subject, using the Fractional Fourier Transform(Fr FT) and Back Propagation(BP) neural network to realize intelligent evaluation and fault diagnosis of switch machines’ work state.In this thesis, FrFT and BP neural network are intergrated to intelligently estimate every switch machine’s current curve of its work state by analysizing the current curves collected from micro-computer monitoring system. Based on the FrFT theory, the signal’s FrFT means a method of expression that coordinate axis rotates arbitrary degree anticlockwise around the origin point in the time frequency plane. The primary action current curves are denoised in the paper for the analysis. And it verifies that the FrFT theory is feasible and advantageous. Then the FrFT theory is used to extract and normalize the feature vector of each type of curve in the intelligent analysis of S700 K switch machine’s action current curves. At last, the normalized feature vectors are difined as the neural network’s system input and the type codes of switch machines’ action current curve are defined as the neural network’s system output. Fault diagnosis system is preliminarily accomplished based on BP neural network to realize the fault intelligent analysis. Based on the condition of neural network’s convergence, the Particle Swarm Optimization(PSO) is used to optimize the training process of neural network to makes it efficient.According to the simulation, it comes to a conclution that the fault diagnosis method of the intelligent analysis of S700 K switch machine’s action current curves based on the FrFT filtering theory and BP neural network is feasible, accurate and intellectualized. It applies to the requirement of the field repair work.