Research On Fault Monitoring And Prediction Based On Kernel Principal Element Analysis And Support Vector Machine

As an important part of mechanical equipment,the safety and reliability of engine operation are directly related to whether the whole set of mechanical equipment can work normally.However,with the development of modern industry,the structure of engine becomes more and the cost has become more expensive.The traditional periodic maintenance strategies can no longer meet the needs of safety and economy.At this time,the development of oil monitoring technology provides a new idea and research direction for the engine from periodic maintenance to condition-based maintenance.However,at present,the analysis of engine lubrication oil data in China is still in the stage of using simple mathematical statistics method for single index.Focing this problem,this thesis combines the physical and chemical indicators of engine lubricating oil and wear debris parameters,then proposes the Kernel Principal Component Analysis Model based on Particle Swarm Optimization(PSO-KPCA Model)and the Support Vector Regression Machine Model based on Particle Swarm Optimization(PSO-SVR Model)for engine fault monitoring and prediction based on the data-driven fault diagnosis technology.Finally,the effectiveness and feasibility of the proposed model are verified by the specific engine lubricating oil data.The model proposed in this thesis makes full use of the various indicators of engine lubricating oil data and improves the reliability of engine fault monitoring and prediction.The specific research content is as follows:Firstly,based on the traditional particle swarm optimization algorithm,this thesis makes two improvements:S-like inertia weight adjustment strategy and mutation operation.This improves the convergence speed and accuracy of particle swarm optimization algorithm,which can be used for parameter optimization of the subsequent proposed model.Secondly,this thesis proposes PSO-KPCA Model for chassis engine fault monitoring,and selects the comprehensive statistics combining T~2 statistics and Q statistics as the monitoring index,which improves the monitoring accuracy of the model and simplifies the subsequent fault prediction model.On this basis,this thesis also proposes a PSO-SVR Model based on comprehensive statistical indicators for the failure prediction of chassis engines.Concerning data,the combination of time series algorithm and support vector regression algorithm are proposed as the theoretical basis of the prediction model.At the same time,in order to make full use of the time series data,the phase space reconstruction method is adopted to extend the one-dimensional data into matrix form,which is helpful to improve the prediction accuracy of the model.Finally,this thesis combines with the actual chassis engine lubrication oil data to test the fault monitoring and prediction model,among which the PSO-KPCA model has a good fault monitoring performance,and the PSO-SVR model can better follow the change of the comprehensive statistics.