Research On Technologies Of Aeroengine Wear Monitoring Based On Lubrication Oil Analysis

Aero-engine as the heart of aircraft supply thrust for it. The operating components ofaero-engine are working in high load, high speed and high temperature chronically, lead tooperating parts are easy to cause failure due to abrasion or fatigue, which have enormousimpact on engine safety work. It is necessary to monitor wear pattern of aero enginecomponents, eliminate hidden dangers in the bud as far as possible. It has significance toguarantee the safety and reliable performance of the engine operation.As the lubricating medium of aero-engine, the lubricant carries important information ofabrasion breakdown. In this paper, firstly, the time sequence sample of lubricant spectrum isanalyzed. The C-C method originating from chaos theory is introduced, and the data samplesare reconstructed with this method. Thus the time delay and dimensions of input samples ofthe model can be determined. Then combining characteristics of spectrum detection data,support vector machine and neural network methods are chose for modeling and predictingthe data sample. Through comparing the simulation results, RBF network model is chose forultimate wear tend prediction.According to the analysis of RBF network learning algorithm, aiming at the effectingparameters of the prediction model, the particle swarm optimization algorithm is improvedand applied on the model optimization. RBF network prediction model of optimum structureis created, and the extrapolation prediction of the aero-engine wearing trend is made. Theinstance simulation results show that, compared with the SVM prediction model, BP networkprediction model and the traditional RBF network prediction model, RBF network predictionmodel, created with the method proposed in this paper, has higher prediction accuracy, andthe prediction results is consistent with the engine actual monitoring data.Finally, based on the tribology theory, the wear pattern and the characteristics of wearparticle are both analyzed, and the wear particle character parameter system is introduced.The recognition model of ferrography wear particle is established and its identification is carried out. Through the instance simulation, the optimum modeling method is chose. Themodel is optimized with the improved particle swarm optimization algorithm, and theself-adapting SVM recognition model is created. Through the simulating experiment on thewear particle, the classification accuracy rate of this method is up to98%. Comparing with theBP network method, the results show that the new method is effective and superior.