Bolt Loosening Detection Based on Principal Component Analysis and Support Vector Machine

Abstract

AbstractOnline monitoring of bolt preload is essential to ensure the proper functioning of bolted structures. Ultrasonic guided wave has the advantages of high sensitivity and wide monitoring range, so it is widely used in the study of bolt loosening monitoring. However, the propagation mechanism of ultrasonic guided waves in bolted connection structure is complicated, and it is difficult to establish a direct relationship between guided wave signal and bolt loosening state directly. In recent years, machine learning and other artificial intelligence technologies have flourished, and a more effective bolt loosening detection technique can be established by using machine learning combined with the principle of guided wave damage detection. In this paper, a bolt loosening identification method based on principal component analysis (PCA) and support vector machine (SVM) is proposed, aiming to achieve end-to-end bolt loosening monitoring with few samples. The ultrasonic wave-guided experimental results of the bolted joint lap plate show that the proposed PCA and SVM technique achieves a loosening recognition accuracy of 92.5%, which is higher than other machine learning methods, and the effects of signal length, number of principal components and the choice of kernel function on the classification performance are explored.KeywordsMachine learningBolt loosening monitoringUltrasound-guided waveSVM