Acta Photonica Sinica, Volume. 50, Issue 12, 1212004(2021)
Debonding Defect Identification Method for Multi-layer Bonded Structures Based on LDA-CPSO-SVM Optimization
Terahertz time domain spectroscopy and support vector machine algorithm are combined to study the defect identification method for multilayer bonded structures. On the one hand, the linear discriminant analysis method was used to reduce the dimension of 14 terahertz time-domain characteristic parameters extracted by the terahertz time-domain spectrum system, and the classification accuracy of normal region, debonding region and edge region in the adhesive layer of multi-layer bonded structure was improved by 20.3%. On the other hand, chaos particle swarm optimization was used to optimize the kernel function of support vector machine, and the classification accuracy of adhesive layer Ⅰ and Ⅱ increased by 18.92% and 9.85% respectively. Linear discriminant analysis based on constructed after parameter optimization of chaotic particle swarm optimization algorithm of support vector machine for multilayer glue joint structure characteristic imaging, the results show that this imaging method can effectively distinguish between sub area of the normal, defect region and edges region, compared with the traditional characteristics of terahertz single imaging technology promoted the debonding defect recognition rate of 50% above,The recognition rate of adhesive layer Ⅰ is 91% and that of adhesive layer Ⅱ is 92%, which greatly improves the recognition ability of debonding defects of multi-layer adhesive structure.
Get Citation
Copy Citation Text
Li ZHENG, Chuang LIU, Jiaojiao REN, Dandan ZHANG, Lijuan LI, Jisheng XU. Debonding Defect Identification Method for Multi-layer Bonded Structures Based on LDA-CPSO-SVM Optimization[J]. Acta Photonica Sinica, 2021, 50(12): 1212004
Category: Instrumentation, Measurement and Metrology
Received: Jun. 25, 2021
Accepted: Aug. 25, 2021
Published Online: Jan. 25, 2022
The Author Email: LIU Chuang (liuchuang@cust.edu.cn)