State recognition of bolted structures based on quasi-analytic wavelet packet transform and generalized Gegenbauer support vector machine

Monitoring the looseness of bolted structures is important to ensure the reliability and integrity of engineering structures. In the past decades, various methods have been developed to characterize looseness states of bolted structures. However, in the long-term storage, transportation, and usage process of bolted structures, especially under random excitation, the vibration-based method for monitoring is the most applicative technology with merits such as low cost and comprehensive operability. To fulfill this task accurately and automatically, a novel looseness state recognition approach of bolted structures based on multi-domain sensitive features derived from quasi-analytic wavelet packet transform (QAWPT) and generalized Gegenbauer support vector machine (GGSVM) is proposed in this paper. To extract effective looseness feature information, the measured non-stationary and nonlinear vibration response signals are processed by QAWPT, and then multi-domain sensitive features are extracted from obtained frequency band signals. For accurate and automatic state recognition, generalized Gegenbauer kernel is introduced, and then multi-class GGSVM is developed to recognize looseness states. In order to validate the effectiveness of the proposed method, a typical bolted beam structure is designed and fabricated, and various looseness states are implemented. The testing results show that the proposed method is effective for looseness state recognition of bolted structures. In addition, QAWPT owns superiority in processing vibration response signals compared with classical WPT, and GGSVM has higher recognition accuracy and better generalization ability than that of kernel SVM with radial basis function.