Research On Intelligent Detection Technology For Aero Engine Borescope Inspection Damage Based On Embedded Device

Aeroengine is the heart of the aircraft,and its structure is precise and complex.It is composed of a large number of components with different functions.The safe and stable linkage of different functional components is the basis for the smooth operation of the aeroengine.During the operation of aeroengine,a large amount of external air will be inhaled,compressed,blended with aviation kerosene,and burned.The resulting expansion gas pushes the turbine to do work,which in turn drives the compressor and the turbofan to rotate.The exhaust gas emitted through the tail nozzle and the rotation of the turbofan jointly generate thrust to propel the aircraft to fly.In this process,the internal components of the aviation engine’s duct not only bear the harsh environment of high temperature,high pressure and high rotated speed,but also bear the impact of foreign matters brought by the inhalation of external air.In combination with the above factors,the internal components of the engine’s duct are prone to burn,cracks,dent,missing coating and other damage,affecting the stable operation of the engine.For this reason,borescope is often used to check the internal damage of the aeroengine through the reserved holes.The appearance of borescope equipment solves the problem of internal damage inspection of the engines.However,there are still some deficiencies in the current borescope inspection,such as fewer professional borescope personnel,low efficiency of borescope defect identification and evaluation,easy to false inspection and missed inspection during borescope inspection,high time cost and economic cost of borescope inspection,etc.Based on this,this thesis studies the intelligent defect identification technology of borescope inspection,and reference on the application of deep learning technology in medicine,driverless vehicles and other fields,it is proposed to use the intelligent damage detection technology of embedded equipment to assist inspector in borescope inspection,improve the detection efficiency,and improve the economic and security benefits of borescope inspection.First of all,this thesis conducts relevant research on the application of deep learning technology in target detection,based on the different mechanisms of prior box generation,analyzes the characteristics of two-stage target detection and single-stage target detection,determines the selected deep learning model YOLOv4,and studies the deep learning framework and target detection performance evaluation indicators.Secondly,this thesis constructs an intelligent detection model for aircraft engine borescope inspection damage,including the establishment of YOLOv4 model environment,data processing and enhancement,data annotation,and model training.In order to ensure better detection results,targeted optimization and improvement work has been carried out the basis of the original model,including clustering the new anchor box using the K-means++ algorithm for better detection head,optimize the activation function of the network and add the Dense Block modules to replace some residual modules in the CSPDarknet53 network in order to obtain better detection accuracy and detection rate.Finally,the embedded terminal equipment is selected,and Jetson Xavier NX is determined as the core of development.The improved YOLOv4 model is deployed on the terminal equipment and Tensor RT inference is performed to improve the detection speed of the model.In order to improve the user experience and carry out friendly human-computer interaction,the external shape of the embedded terminal was designed,the interaction software was developed and relevant tests were carried out.The tests indicate that the device can be integrated with the boreoscope inspection to perform real-time processing of the image data input from the boreoscope inspection devices,predict the type of damage in the image data,and indicate the damaged area.Additionally,it can be used separately for rechecking the boreoscope inspection video,greatly improving the efficiency of engine boreoscope damage detection and reinspection,and has a promising application prospect.