Automatic detection method for loose bolts in steel frame structures based on wearable lightweight impedance system and multi-channel information fusion deep learning model

Bolt connection is widely utilized in small and medium-sized engineering structures. However, due to the cumulative effects of environmental factors, loading conditions, and other variables, bolts are prone to loosening, posing significant threats to structural safety. The objective of this research is to develop an efficient and accurate automatic detection method for bolt loosening in steel frame structures. To achieve this, a novel approach integrating a wearable lightweight impedance system and a deep learning model that fuses multi-source information is proposed. A wearable impedance monitoring system with a multi-channel wearable washer and a miniature impedance assessment board is designed, eliminating the need for pre-pasting sensors and overcoming the limitations of traditional impedance monitoring devices. Additionally, a signal transformation method converting one-dimensional signals into multivariate recursive graphs and extracting recursive characteristics is developed to better represent bolt-loosening states. The innovative MonitTransformer model is then proposed to process these graphs, automatically identifying bolt loosening states without manual feature engineering. Experimental results on steel frame structures indicate that the wearable impedance monitoring system can effectively detect bolt loosening, and the MonitTransformer model outperforms existing methods in recognition accuracy. This research provides a practical solution for real-time monitoring and rapid assessment of large-scale bolt loosening in IoT-based structural health monitoring, demonstrating great potential for enhancing the safety and reliability of engineering structures.