INTELLIGENT DETECTION OF BOLT LOOSENESS USING STRUCTURAL HEALTH MONITORING METHODS AND PERCUSSION APPROACH
Date
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Bolted joints have been widely used to connect different components across multiple engineering fields, while the bolt looseness detection is an urgent issue to be solved. Recently, several piezo-enabled structural health monitoring (SHM) methods have been utilized to detect bolt looseness, including the active sensing method, electromechanical impedance (EMI) method, and the vibro-acoustic modulation (VAM) method. However, current approaches mostly focus on single-bolt looseness detection, and there is still a lack theoretical investigation to explore the principle of these methods. In this dissertation, several in-depth studies to advance the development of research in the bolt looseness detection are presented. First, a numerical model, a semi-analytical model, and an analytical model of the active sensing method for single-bolt looseness detection is proposed. Then, several new entropy-based indices are developed to replace the current index, i.e., signal energy. Via these entropy-based indices and machine learning (ML) technique, the detection of multi-bolt looseness is achieved for the first time. Second, a model to describe the relationship between bolt preload and EMI signal is theoretically developed, providing a better understanding of the EMI method. Third, in terms of the VAM method, swept sine waves as inputs are employed to improve practicability, and a new entropy-based index is developed to enable the VAM method to detect multi-bolt looseness. Moreover, considering that the above methods depend on permanent contact between transducers and structures, a new percussion-based approach is proposed. By tapping the bolted joint and analyzing the percussion-induced sound signals, the bolt looseness can be detected without contact-type sensors. First, an analytical model to research the mechanism of the percussion-based approach for bolt looseness detection is proposed. Then, by using deep learning (DL) based techniques to process and classify the percussion-induced sound signals under different bolt preloads, two practical percussion-based approaches to detect bolt looseness detection were developed. In summary, several in-depth investigations of SHM methods and a new percussion-based approach for bolt looseness detection have been conducted in this dissertation. It is believed that these methods have great potential for future industrial applications.