Research On Data-driven Fault Diagnosistechnology Of Large Aircraft PDU Component

With the increasingly important role of airfreight in China's economic production and life,the mechanization,automation and intelligence of cargo loading systems have become an irreversible trend,resulting in the maintainability demand of the loading system and its key components has become more apparent.The cargo loading system contains various electronic and mechanical subsystems and key components.The complexity of the system is high and the possibility of system failure increases accordingly.Once the fault occurs,the chain reaction may lead to the system extension and even destructive consequences,resulting in unnecessary economic losses.The PDU is a power drive unit in the loading system.It plays an important role in the system and is prone to failure.Therefore,fault diagnosis for PDUs has become an important subject to be studied and solved.With the development of fault diagnosis technology,along with the continuous research of machine learning and data mining technology,the value of data for fault diagnosis has become increasingly prominent.How to effectively use existing data of the equipment to deeply mine the valuable fault information contained in it becomes an important research content of the data-driven fault diagnosis method.In this paper,the data-driven fault diagnosis method is researched and improved mainly from the aspects of feature extraction method and fault diagnosis method.The validity and superiority of the method are verified by theoretical derivation and TE dataset simulation.Combined with the public information of PDU and the existing PDU maintenance experience,the PDU failure mechanism and condition monitoring method are explored,which lays a foundation for further obtaining PDU measured data and realizing PDU fault diagnosis.The main research results of this paper are:1.The median-based WPCA algorithm feature extraction method is proposed,which solves the problem of high difficulty,poor extraction and poor performance due to the high dimensional,multi-redundancy and noise-containing characteristics of the original data.The TE dataset is used to simulate and verify the algorithm.The results show that the proposed feature extraction method improves the performance of traditional PCA under the premise of ensuring the efficient operation of the algorithm.The processing results can highlight the main features or main dimensions of the original data,and reduce the impactof secondary features and noise signals.There are two main improvements:1)By calculating the weighting of each dimension feature for the different contribution degree of the classification,more accurate and effective feature extraction is realized,and the weighted processing result can reflect the main characteristics of the data.2)In the implementation method of the traditional PCA centralization preprocessing,the median value that is not affected by the extreme values of the data is used instead of the mean value,which avoids the extracted features from being affected by the extreme values of the original data(i.e.noise and abnormal signals).Feature extraction results become more accurate and reliable.2.An improved fault diagnosis method based on clustering algorithm—the clustering map method of diagnosis set is proposed,which solves the problem that the existing fault diagnosis method is generally more complicated.By constructing a standard feature sample set and constructing a diagnostic set with the sample to be diagnosed as the input of the clustering algorithm,the visual representation of the fault detection and fault diagnosis results on the clustering result map is realized.At the same time,the method is combined with the median-based WPCA algorithm feature extraction method to improve the clustering effect and the correct rate of fault diagnosis.3.The PDU failure mechanism and condition monitoring method are explored,and the problem of the correspondence between the PDU failure types and the failure symptoms is solved.A method is given for solving the implicit mapping relationship between various faults and dynamic parameters by kinetic modeling.A preliminary PDU condition monitoring scheme is designed.It lays a foundation for obtaining the measured status data of the PDU and further realizing the fault diagnosis of the PDU.