Research On Real-time Detection Technology Of Solid Rocket Motor Grain Defects Based On Dual View CT Reconstruction

The health status of solid rocket motor is one of the important factors for the success of the mission.At present,the nondestructive testing technology for ground testing of engines has been relatively mature,but during the ignition test,the defects such as cracks produced by the grain due to the load will also cause the launch failure,so it is necessary to complete the realtime detection of defects under thermal conditions.However,the traditional CT reconstruction algorithm and the existing deep learning method are difficult to reconstruct defects in real time and effectively under the condition of ultra-sparse acquisition conditions and without defect prior information,and can not detect defects quickly and accurately when the engine is working.In order to solve the above problems,this paper uses deep learning technology combined with the principle of parallax to carry out the research on the real-time detection algorithm of column defects under the ignition test of solid engine.In order to complete the real-time detection of grain defects in the solid rocket motor ignition test,this paper designs a mutually orthogonal dual-view imaging system,and studies the real-time detection algorithm of grain defects in the motor based on dual-view CT reconstruction.Firstly,the deep learning model is used to reconstruct the internal structure of the engine from two angles of view,and the mapping from 2D to 3D is completed;Then,the unsupervised anomaly detection algorithm is used to detect and locate defects in the dual-view image;Finally,the principle of parallax is used to reconstruct defects in 3D structure to achieve the purpose of dynamic monitoring.However,considering the poor generalization ability of the anomaly detection network,and the low detection accuracy and insufficient image matching corners,it is difficult to achieve fine characterization of defects.Therefore,this paper studies the grain defect detection algorithm based on the data enhancement strategy,uses the generative adversarial networks to enhance the data,and improves the anomaly detection network in combination with the characteristics of the grain RTR image,effectively improving the generalization ability and detection accuracy.In addition,the field of view conversion model is used to find the corresponding matching points of the dual-view image to further realize the accurate reconstruction of defects.The experimental results show that the proposed algorithm can effectively complete 3D reconstruction for defect detection only by using the projected images from two perspectives.Compared with the existing ultra-sparse reconstruction algorithm based on deep learning,it does not need the prior information of the defect area,and provides a way to detect the grain defect in the case of solid rocket motor ignition.