Aircraft Engine Fault Diagnosis Based On Twin Support Vector Machines

Aircraft engine fault diagnosis technology is an important supporting technology in the transformation process of aircraft maintenance.The advanced fault diagnosis algorithms determine the degree of advanced fault diagnosis technology of aircraft engine greatly.So the performance of fault diagnosis algorithms directly affects process of aircraft maintenance reform.In this thesis,fault diagnosis of aircraft engine is studied based on the twin support vector machine and its improved algorithm.This thesis includes specifically:Firstly the twin support vector machine classification algorithm is used to research on aircraft engine fault diagnosis.The twin support vector machine is a new achievement of the support vector machine algorithm,which has faster computing speed and other advantages.In this process,a mixed kernel function is adopted to complete nonlinear classification,because it could better balance between learning ability and generalization ability.The hybrid particle swarm optimization is adopted to optimize parameters of the nonlinear classification model because of its better performance in finding the global optimal point.Results of simulation reveal that this algorithm could get an excellent result of fault classification.Secondly the least square twin support vector machine classification algorithm is introduced into aircraft fault diagnosis.Least squares twin support vector machine is a least squares form of the twin support vector machine algorithm,which can improve speed of this algorithm.A novel binary tree multi-classification algorithm is proposed based on the “barycenter” concept for building multi-classification model,which could reduce calculated amount to a certain extent.Results of simulation reveal that this algorithm could get an excellent result of fault classification.Thirdly the regularized twin support vector machine has a more complete theory and has better classification performance.So this algorithm is used to do aircraft engine fault diagnosis.Adaptive decision directed acyclic graph classification algorithm is improved from decision directed acyclic graph algorithm,which can reduce number of layer to reduce "error accumulation",so this algorithm is applied to multi-classification algorithm of regularization twin support vector machine.A suitable parameter optimization algorithm is used to optimize the aircraft engine fault diagnosis model.Results of simulation reveal that this algorithm could get an excellent result of fault classification.Finally the least squares twin parametric-margin support vector machine is the latest achievement of parametric-margin support vector machines.Its unique algorithm design idea has its own unique advantages.In this thesis,this algorithm is used in fault diagnosis field,a suitable multi-classification algorithm is designed,and the particle swarm optimization algorithm is use to find the best parameters of model.Results of simulation reveal that this algorithm could get an excellent result of fault classification.