Research And Application Of Aviation Material Parts Query System Based On Deep Learning

The query of aviation material parts is an important step of the work of aircraft maintenance personnel,and its query accuracy has a direct impact on aircraft flight safety.Through practical investigation,it is known that the current parts query is mainly based on the Illustrated Parts Catalog of the aircraft type of parts to artificial search.This query method has the problems of slow query speed and low accuracy,and it is often difficult to meet the demand of the part number in actual maintenance work.In order to solve the above problems,the SR20 aircraft IO360ES engine parts are taken as the research object.By constructing the image sample set of aviation material parts,the effective features of parts are extracted by using the deep learning algorithm,the automatic query of part number information is realized,and the query system of aviation material parts is designed and deployed.The main work is as follows:Firstly,the mechanism of deep neural network is discussed in this thesis,including neuron structure and perceptron network.In this thesis,the convolution,pooling,and full connection in the convolution neural network structure are analyzed in detail.The important indicators such as Intersection over Union and Non-Maximum Suppression in the basis of image detection network are expounded,which provides theoretical support for the subsequent construction of the parts query model.Secondly,the collection requirements,collection methods and classification of SR20 engine parts are analyzed,and the collected sample data of aviation material parts are cleaned and repaired by image.The data sample set of aviation material parts is constructed by image enhancement and image annotation,and the features of parts are extracted by deep learning algorithm,which provides data basis for the training of deep neural network in the next chapter.Thirdly,the YOLOv5-F component query model is built based on the sample data of aviation material parts.The prediction model that meets the design requirements is obtained by multiple iterative training of the model using the constructed sample set of aviation material parts.Comparison between prediction model and other detection algorithms through query experiments,and the comparison results show that the optimized network model has a great improvement in the accuracy of parts query.Finally,the Qt Designer software is used to visual design the program of aviation material parts query system,and the relevant query platform is built.The system is utilized to query the aviation material part number for the parts images,offline parts videos and online parts videos,and the part number query result can be displayed on the front-end interface simultaneously.Compared with other query methods,it is proved that the system can quickly and accurately query the aviation material part number information.