Comparison and fusion of various classification methods applied to aero-engine fault diagnosis

Aero-engine fault diagnosis plays a crucial role in safe operation and cost-effective maintenance. Early detection and isolation of component faults prior to failure of aero-engines is of utmost importance. This paper applied various classification methods, including Support Vector Machine (SVM), Decision Tree (DT), K-Nearest Neighbors (K-NN) and Linear Discriminant Analysis (LDA), to aero-engine component fault detection and isolation. These 4 various methods were tested under different circumstances involving large training samples, small training samples, few features, and noisy data. Comparison of accuracy and time efficiency was made among those methods, and a fusion algorithm of these 4 classification methods was proposed based on their training classification accuracy. The experiment results show that SVM has excellent accuracy performance but poor time efficiency while LDA performs well in both accuracy and time efficiency, and the fusion algorithm has higher accuracy than any single method.