| Carbon fiber composite materials have many advantages such as low friction coefficient,good fatigue resistance,high specific strength and specific rigidity,strong corrosion resistance and wear resistance,etc.,and are widely used in spacecraft such as aircraft,satellites,and space particle detectors.However,during the production,manufacturing,processing and long-term service,the composite material components will be affected by various external factors,resulting in defects such as delamination,debonding,cracks,and inclusions,which will further damage the spacecraft and cause serious losses.Therefore,the defect detection of spacecraft composite materials has important research significance.This paper takes the X-ray inspection images of spacecraft composite materials as the research goal,and proposes a defect detection algorithm based on lightweight convolutional neural network to study the defect detection of spacecraft composite materials.Firstly,according to the technica l requirements of the spacecraft composite material defect detection system,the overall solution is designed.The X-ray image of the composite material with delamination,debonding,crack,inclusion,voids and other defects was obtained by X-ray inspection system.The X-ray detection system provides sample data for the training and detection of defect detection algorithm model proposed later.Secondly,the lightweight method of convolutional neural network and the model of target detection network are stud ied,and on the basis of these methods,the defect detection algorithm based on lightweight convolutional neural network is proposed.Based on the FCOS model,the algorithm replaces the FCOS feature extraction network with a lightweight network that combines the advantages of SqueezeNext block and IBN-Net,which further enhances the generalization ability of the algorithm.In addition,a defect data set of spacecraft composite materials is constructed,and several data expansion methods are proposed to solve the problem of insufficient data samples.Thirdly,the software for the defect detection system of spacecraft composite materials is designed and developed.The system software uses a cross-platform graphical interface development framework Qt to achieve user login,defect detection,defect data storage and management,and test report generation.By replacing the data set and retraining the model,the software can also detect defects in other materials.Finally,the defect detection algorithm and defect d etection system proposed in this paper are tested and analyzed.At the same time,in order to verify the performance of the defect detection algorithm in this article,an experiment is conducted on the embedded platform NVIDIAJetson TX2.By setting the highest power mode,only0.26 s of detection time can achieve 90.3% detection accuracy. |
