The Research On Fault Diagnosis Technology Of Electric Locomotive Main Transformer

Abstract:Under the theme of "high-speed passenger, freight overloading", China Railway career has achieved leapfrog development in the21st century. Meanwhile, with the acceleration of train operation speed, the increase of the single locomotive traction tonnage, the higher requirements are put forward for locomotive and vehicle’s repair and maintenance. As the energy source of electric locomotive, the main transformer is the heart of the electric locomotive and plays an important role in guaranteeing railway transport’s safety and efficiency. However, in contrast to the importance of main transformer in the electric locomotive, the research on the fault diagnosis technology of main transformer is obviously not enough. Therefore, the fault diagnosis research of electric locomotive main transformer has great significance in improving the locomotive overhaul security level, perfecting the theory of electric locomotive fault diagnosis technology and enhancing the running safety and reliability of electric locomotive/emus.The study object of my dissertation is the main transformer of electric locomotive. On the basis of summarizing the structure and application characteristics of the main transformer in electric locomotive, this paper researches and analyses the characteristics of the vibration signals of electric locomotive main transformer tank and dissolved gas in transformer from the aspects of mechanical vibration, electromagnetic, electrochemical and electrical science.And the fault diagnosis technology of the main transformer in electric locomotive,which meets the actual demand of our country current electric locomotive main transformer repair and need of the developing locomotive "status maintenance" currently, is deeply exploredAiming at the Distortion Failure of Locomotive transformer winding and core, this paper presents a new fault diagnosis technology, which is based on the vibration of electric locomotive transformer tank. The technology is not only higher sensitivity, but also could provide a basis for the development of the main transformer online condition monitoring. Firstly, this article analyzes the vibration signal source of locomotive main transformer detailly from the transformer vibration mechanism, and discusses how to choose the signal measurement location of locomotive main transformer tank analyzed. Then, on the one hand, the paper establishes the equivalent mathematical model for the winding vibration of locomotive main transformer by using quality-spring model, and deduces the winding vibration acceleration equation when the main transformer works steadily; on the other hand, it also deeply discusses the main cause of transformer iron core vibration and the factors that influencing the vibration signal characteristics.Meanwhile, for the fault diagnosis of winding deformation of locomotive main transformer, the dissertation obtains transformer’s electromagnetic field coupling equations through the Maxwell equation and its equivalent circuit, and builds the solid finite element model for HXD1C locomotive vehicles main transformer by ANASYS software. Through working on this model, the variation characteristics of winding vibration signals are studied under different prestressing force, and a monitoring method of transformer winding pre-tightening force based on100HZ winding axial vibration signal is proposed here. According to the signal characteristics of the transformer core vibration, this paper proposes a training algorithm of wavelet neural network based on hybrid particle swarm optimization algorithm, which is used to diagnose the looseness fault of electric locomotive traction transformer iron core. The MATLAB simulation tests show that the wavelet neural network based on this optimization algorithm has faster convergence speed and higher precision on diagnosing the looseness of the electric locomotive traction transformer iron core by vibration signal.Aiming at the problems encountered during the process of DGA technology applied on the fault diagnosis for electric locomotive main transformer. After systematically analyzing the principle and the existing diagnosis algorithm of DGA, the article proposes a kind of self-organizing RBF neural network training algorithm to be applied in electric locomotive main transformer DGA fault diagnosis technology, by organically integrating a variety of DGA diagnosis method and combining with the characteristics of locomotive main transformer. In order to enhance the global search ability of traditional PSO, the proposed algorithm uses Average Seed Spacing to describe particles’concentration level, and increases the inertial factor of PSO algorithm in according to certain probability,which combines with the Gaussian random number; in the meantime, for the purpose of ameliorating the deficiency of past RBF neural networks, the FCM algorithm and Gaussian-PSO algorithm is applied to RBF neural network to select hidden layer nodes and optimize the network connection weights. The iris data set and wine set algorithm are used to verified it. MATLAB simulation test show that the algorithm has higher diagnosis accuracy is indeed, but the training time is longer.Finally, in view of the backwardness of test equipment for the locomotive main transformer maintenance, after detailly studying on the basic requirements of locomotive main transformer type testing and the existing main problems in locomotive main transformer maintenance and repair work, the paper presents the software and hardware design of main transformer comprehensive test and fault diagnosis system. The system can meet all type tests requirements of domestic mainstream electric locomotive main transformer and can diagnose the tested transformer faults comprehensively by using the dissolved gas in transformer oil, transformer tank vibration signal and type test data.