Research On Pulsed Infrared Thermal Wave Detection Technology For CFRP Laminate Defects

Carbon Fiber Reinforced Polymer(CFRP)is one of the most widely used and most mature structural composites.Because of its good corrosion resistance and heat resistance,low thermal expansion coefficient,high specific strength and high specific modulus,it is widely used in aerospace,military industry,electronics and other fields.In the process of preparation,processing and service of CFRP laminates,due to the influence of manufacturing process,fatigue,cyclic stress and impact damage,it is easy to form defects such as delamination and debonding between layers,which seriously affect the quality and performance of CFRP laminates,and there are huge potential safety hazards.Therefore,it is necessary to carry out reasonable and effective detection and evaluation of carbon fiber reinforced composite components.This paper mainly focuses on the theory and simulation of pulsed infrared thermal wave imaging detection of delamination defects in CFRP laminates,experimental research,infrared thermal image sequence processing,CFRP laminates defect identification and segmentation,CFRP laminates defect area and depth prediction and quantitative analysis,etc.,to provide a powerful supplement for the quantitative evaluation of CFRP laminates defects.According to the related theories of material science and heat transfer,a three-dimensional heat conduction mathematical model of pulsed heat flow in CFRP laminates was established,and the finite element method was used for simulation research.The relationship between the geometric characteristics of the defect and the distribution of the surface temperature field,and the relationship between the detection process parameters and the defect recognition effect are studied and analyzed,which provides a theoretical basis for the experimental research.A pulsed infrared thermal wave imaging detection system was built,and the delamination defects of CFRP laminates were successfully detected.The influence of detection process parameters and defect plane size and depth on the detection effect was explored.At the same time,the experimental results are compared with the simulation results to verify the accuracy of the heat transfer model and obtain the appropriate detection process parameter range.The image sequence is preprocessed by removing the fitting background sequence method and the median filtering method,which effectively solves the problem of thermal uneven noise and speckle noise.The inter-frame difference-multi-frame cumulative average method,polynomial fitting method and principal component analysis method are employed to extract and process features from sequences of infrared thermal images,and the signal-to-noise ratio and structural similarity are introduced to quantitatively evaluate the processing effect of each algorithm.The results show that the principal component analysis method has the best processing effect,and the obtained image has strong defect recognition ability and good structural similarity.The research on defect recognition and segmentation of CFRP laminates,defect area prediction and quantitative analysis of CFRP laminates was carried out.A semantic segmentation algorithm based on Mobile Net V2 and Deep Lab V3+was proposed to segment the input image and quantitatively calculate the defect area,and compared with the fuzzy C-means clustering-dynamic threshold segmentation algorithm.The results show that the semantic segmentation algorithm based on Mobile Net V2 and Deep Lab V3+has a good effect on defect segmentation,and the prediction error of defect area is small.The prediction error is between 0.4%and 32.5%,which can realize the prediction of defect area of CFRP laminates.The quantitative analysis of the defect depth of CFRP laminates was carried out.Four defect depth prediction models,PSO-LSSVM,BPNN,PSO-BPNN and WOA-BPNN,were proposed to realize the effective identification of CFRP laminates defects and the prediction of defect depth.MAE,RMSE and R~2 were introduced to evaluate the prediction results.The results show that WOA-BPNN has higher prediction accuracy and goodness of fit.The defect depth predicted by WOA-BPNN is quantitatively analyzed.According to the analysis results,the defect depth prediction error of WOA-BPNN is maintained within 5%,and the minimum prediction error is only 0.3%.The prediction effect is good.