Research On Aeroengine Life Prognostics Based On Information Fusion Technologies

The working environment of aeroengine is harsh,which is easy to cause serious catastrophic accidents,so it is importsnt to predict the useful life of aeroengine to improve its reliability and safety.The research is mainly focused on the prediction of remaining service life(RUL)of aeroengine parts or the whole equipment.Based on information fusion techiniques,this paper focus on the prediction of remaining service life of aeroengine whole equipment.The main work is as follows:Firstly,the aeroengine health index is calculated based on the idea of information fusion in data layer.It is mainly carried out in three steps: selection of sensor data,data preprocessing and establishment of health assessment model.The selection of sensors is based on the observation and analysis of the change trend of all sensor data in the time series.The data preprocessing includes data drying and data normalization.In the health assessment model,Kernel principal component analysis(KPCA)method is used to fuse the preprocessed sensor data in the data layer.The health index(HI)value which can reflect the overall performance degradation of the engine is obtained.The trend chart shows that the HI value can reflect the performance degradation trend of the equipment better than the single sensor data,and also better reflect the health status of the aeroengine.Secondly,three prediction models based on data-driven are selected to predict the RUL of aeroengine,and the performance is compared and analyzed.Firstly,the Euclidean distance is selected as the measure function in the trajectory similarity based prognostic(TSBP)prediction model,and the evaluation,decoding and training problems are solved in the hidden semi-markov model(HSMM)prediction model.The HI value obtained from data fusion is applied to these two prediction models.Quantum particle swarm optimization(QPSO)is used to optimize the parameters of support vector regression(SVR)model.Then,three single prediction models are used to predict the RUL of aeroengine.Finally,the three prediction results are compared and analyzed.It is concluded that theprediction performance of QPSO-SVR algorithm and HSMM algorithm is better when the RUL of aeroengine is less(equipment degradation approaching failure),and the prediction performance of TSBP is better when the RUL of aeroengine is more(equipment performance status is normal or degradation just occurs).Finally,in order to improve the robustness and accuracy of RUL prediction results,decision information fusion is carried out for the prediction results of the above three prediction models.Three methods,precision weighted,diversity weighted and information entropy,are mainly used for decision-making level fusion,and the results of three kinds of decision-making fusion are compared.the prediction results of decision-making level fusion are compared with those of TSBP,HSMM and QPSO-SVR.The results show that the prediction results after the fusion of decision-making level are better than those of single prediction model,whether in the stage of less RUL(equipment degradation approaching failure)or more RUL(equipment performance state is normal or degradation just occurs).