Debonding Detection in Carbon Fiber Reinforced Polymer Plate Repaired Steel Beam Using Percussion and Gaussian Mixture Model Clustering
Debonding Detection in Carbon Fiber Reinforced Polymer Plate Repaired Steel Beam Using Percussion and Gaussian Mixture Model Clustering
复制标题
DOI：
10.1109/iccsi55536.2022.9970670
复制
发表时间：
2022-11
期刊：
影响因子：
--
通讯作者：
Yong Xu;Ji-An Chen;Xuemin Chen;Gangbing Song
中科院分区：
文献类型：
--
作者：
Yong Xu;Ji-An Chen;Xuemin Chen;Gangbing Song
碳纤维增强聚合物（CFRP）已被证明是一种经济有效、高效且可靠的结构修复或加固方法。脱粘检测是确保此类修复的完整性和性能的重要措施。本文提出了一种利用冲击和无监督机器学习检测CFRP板修复钢结构脱粘的方法。使用具有粘结 CFRP 和已知粘结缺陷的钢梁作为测试样本。然后，轻敲横梁上不同粘合条件的不同位置，产生敲击声，并由 iPhone 记录下来。采用梅尔频率倒谱系数（MFCC）算法从打击乐声音中提取特征。采用无监督机器学习算法、高斯混合模型 (GMM) 聚类来进行脱粘检测。该方法对测试样品的脱粘检测准确率达到95.8%。
The carbon fiber reinforced polymer (CFRP) has been proven to be a cost-effective, efficient, and reliable method for structural rehabilitation or reinforcement. Debonding detection is an important measure to ensure the integrity and performance of such repairs. In this paper, a method of using percussion and unsupervised machine learning to detect the debonding of CFRP plate repaired steel structure is proposed. A steel beam with bonded CFRP and known bonding defects is used as a test specimen. Then, different locations with different bonding conditions on the beam are tapped to generate the percussion sounds, which are recorded by an iPhone. The mel-frequency cepstral coefficient (MFCC) algorithm is employed to extract features from percussion sounds. The unsupervised machine learning algorithm, Gaussian mixture model (GMM) clustering, is implemented for debonding detection. The proposed method achieves 95.8% accuracy in debonding detection on the test specimen.