Research On Aero-engine Fault Diagnosis Base On Case-Based Reasoning And Improved Neural Network

The fault diagnosis of aero-engine has always been the Research focus of civil aviation field and has very important significance for the airlines' safety. If can identify potential faults rapidly and accurately through analyzing the change of engine monitoring parameters, the airlines can make the maintenance plan better, shorten the troubleshooting time, and determine the range of work in the engine repairs and depth, which can not only reduce the maintenance time, operation cost, and increase the time on-wing of engine, so as to improve the overall benefits of the airline, but also effectively avoid in-flight shut down and flight delays caused by the fault of the engine.The research has collect the actual fault cases of aero-engine, and mainly studied the integration of CBR and neural network method of engine fault diagnosis, the main research content:1. CBR method faced the problem of lacking attribute parameters and non-ideal diagnosis result of engine fault diagnosis. The research proposes a way to analysis the response of monitoring parameters in some critical points after failure by using QAR data, and compared to the typical fault cases, this method extension the fault cases attributes parameter index, which laid a solid foundation for improving the diagnosis accuracy.2. The difficult in building the engine fault case base is that the rules is hard to find and the low efficiency of retrieval algorithm. The paper solve the problem by building the structure of the engine fault case information, and then build the case base which can satisfy the demand of engine fault diagnosis; The Sigmoid function and WFA attribute weight allocation method has been introduced to improve the grey correlation algorithm,which makes the attribute weights can be adjusted automatically according to its values,the calculation results show that the matching accuracy is also improved.3. The paper designed the fault classifier model base on PSO algorithm optimizes the BP network structure, the results show that PSO-BP algorithm has higher training accuracy and learning ability compared to the effect of engine fault diagnosis of GA-BP network and using the Iris data set for classification experiment, which help improve the calculation precision and speed in engine fault diagnosis.4. The research designed the engine fault diagnosis system platform based on CBR-NN model by combining CBR method and neural network. The actual case result of diagnosis shows that the system can minimize misjudgment, and help the engine monitor engineers analysis the QAR data and diagnosis fault effectively, which has the important value for engineering application.