Fault Diagnosis Method Based on LSA and SVM

Against the application problems of support vector machine in fault diagnosis, the paper introduces a new diagnosis method, which combines latent semantic analysis with improved support vector machine. With latent semantic analysis realizing sample datum feature extraction and dimensionality reduction to solve the training and diagnosing speed problem, resulted from a number of high-dimensional sample datum; and with the improved max-wins-voting strategy one-versus-one classification to solve the unclassifiable problem in conventional method, improving the efficiency and accuracy of fault diagnosis. Applying the above method to fault diagnosis for a certain type of missile, the accuracy could reach to more than 94%, and the experimental results demonstrates the superiority of the presented method and its applied value.