Research On Fault Diagnosis Of Batch Process And Condition Monitoring Of Aircraft Steering Engine

With the continuous progress of modern industry technology,the security and reliability of the system has created higher requirements,so fault diagnosis technology has also developed.The aircraft steering engine is the core executive mechanism of civil aircraft flight control system,which drives the deflection of control surfaces such as elevator,aileron and rudder to realize the control of aircraft attitudes and trajectories.Because of its complex structure of non-linear closed-loop system,aircraft steering engine belongs to fault-prone system.While the operation of civil aviation requires high safety and reliability of airborne components,so it is of great significance to study the technology of batch process fault diagnosis and aircraft steering engine condition monitoring data-driven-based algorithm.Based on the existing fault diagnosis technology,taking the aircraft steering engine system as the research object,this thesis systematically explores the process-based multivariate statistical analysis methods from three aspects: affinity propagation cluster analysis,principal component analysis and multi-way principal component analysis,and carries out the following independent work.A two-step method based on improved affinity propagation(AP)clustering and sub-phase similarity diminishing scan(PSDS)is proposed to model and monitor batch process data.First,to capture the dynamic characteristics of modes switching,the classical AP clustering is improved by using similarity measure method which can reflect the trend of data characteristics more accurately in complex batch process.Then the improved AP clustering algorithm is adopted to conduct the phase preliminary partition.Considering that the internal features of the sub-phase obtained from the phase preliminary partition also shows significant trend of change,the PSDS method is then proposed to implement phase fine partition.Each sub-phase scanned by the PSDS method is identified and divided into stable parts and transition parts.Meanwhile,the concepts of outliers and misclassification points and their solutions are put forward in the process of modeling.For some problems that commonly exist in fault diagnosis of batch process,such as timesequence,uneven-duration with out-of-sync trajectories,this thesis further proposes a multi-phase process monitoring scheme based on sequential moving principal component analysis(SMPCA).First,the SMPCA is proposed to perform the multi-phase partition of uneven-duration batches in time sequence.The update of the SMPCA exactly explains the dynamic characteristics of the sampling data.Subsequently,the essential trend of process change is also explained by extracting the local SMPCA models' feature space and establishing the similarity evolution index for critical variables.Finally,the modular modeling is conducted to solve the problem of uneven-duration batches with out-of-sync trajectories.Then a monitoring technique including the differentiated monitoring and secondary monitoring is introduced to improve the overall monitoring performance of the model.To verify the performance of the proposed methods,this thesis adopts the sampling data of the aircraft steering engine to carry out condition monitoring.Through experiments and comparative analysis,the proposed algorithm has more reliable condition monitoring performance.