基于机器学习的黏钢构件黏接层缺陷识别方法研究OA北大核心CSTPCD

The effects of bonding layer defects on ultrasonic detection signals of bonded steel reinforced struc-tures were deeply studied and a new method for the bonding layer defect identification based on machine learn-ing was proposed.Firstly,based on the direct contact pulse-echo reflection method,the finite element simula-tion of the viscous steel member was carried out,and the propagation law of ultrasonic waves in the viscous steel member was expounded.Secondly,the characteristics of local ultrasonic echo signals and related signals were analyzed,and the effects of different defect variables on ultrasonic echo signals were discussed.Finally,the ultrasonic time-history response data set of the adhesive steel member was established,and the classifica-tion and recognition performances of different machine learning models for the size and location of defects were compared,and the defect identification method for the adhesive layer of the bonded steel member was built.The results show that,the local ultrasonic echo signal and its characteristics change regularly with the defect size and location,which can help preliminarily distinguish the defect information.Meanwhile,the proposed RF model-based defect identification method can effectively identify the defects of the adhesive layer in the bonded steel member,and has a broad engineering application prospect.