Research On Fault Diagnosis Of Aero-engine Based On Neural Network

A method based on Neural Network for fault diagnosis of aero-engine is put forward. Thispaper contains the description of the structure characteristics and working principle ofaero-engine, and the list of different engine faults. We make the connection between faultmodes and data characteristics by analyzing the state information of the aero-engine, andbuild the fault mode structure; depending on expert experience and experiments, the faultmode database that contains the trends of characteristics in different fault modes and used forexperiments, is built; the Neural Network is applied to construct the intelligent fault diagnosismodel, as well as classify and define the faults. Principal Component Analysis(PCA) is usedfor the purpose of feature extraction and dimension reduction before diagnosis process tomake the computation more effective in high-dimension condition. The data characteristicsthat got through the PCA, as the neural network input, are used for training and test. MLP,RBF and SOM network are applied to build the fault diagnosis model to realize faultclassification. Because of the diversity of neural network structures, the fault diagnosismethods are different, and we design the diagnosis algorithm separately. The experimentsshow that the diagnosis methods are feasible. The fault diagnosis method, Neural Networkwith PCA data preprocess, improves the work clearly, and makes a good performance that thediagnosis correct rates stay at a high level in application. It’s with highly practical value infault diagnosis area.