Research On Fault Diagnosis Method Of Aircraft Generator Rotating Rectifier Based On ELM

With the proposal and development of the concept of more-electric and all-electric-aircraft,Aircraft electrical power system become more and more important. People also paid more attention in the reliability, maintainability and testability of the system. Aircraft generator is the main power supply. As the core component of aircraft electrical power system, whether it works normally affects the power supply capability directly and the flight state of the aircraft furtherly. Rotating rectifier is the key component of the aircraft generator, therefore, research on fault diagnosis method of aircraft generator rotating rectifier is of great significance to improve the operational reliability of the aircraft generator and guarantee the safe operation of the aircraft.This paper focuses on researching of fault diagnosis method of aircraft generator rotating rectifier based on Extreme Learning Machine(or ELM), and concrete research content is as follows:(1) Establish the overall model of aircraft generator through the Matlab/Simulink software,Analyze the correlative signal of the various fault modes of rotating rectifier and lay the foundation for subsequent fault diagnosis.(2) Research on the basic principle of extreme learning machine algorithm, and two improved methods were put forward according to some limitations of ELM which respectively based on Mind Evolutionary Algorithm and Teaching-Learning-Based Optimization. ELM and two improved methods were utilized to classify the simulation data of aircraft generator model, and then compared with the BP neural network and support vector machine. The experimental results are analyzed lastly.(3) The generator fault simulation test platform was set up, on which related fault experiments were carried out. The measured data was collected by setting the rotating rectifier fault when the generator supplies no-load and balanced resistive loads. A variety of fault diagnosis methods were implemented, the performance of which were compared, analyzed and summarized.