Advanced Structural Health Monitoring for Connections with PZT Transducers and Image Processing Technique

Mechanical connections, including nut and bolt connections, welds, screw joints, pin connections, among others, are used in every industry. As connections are deployed into a variety of environments, a multitude of factors may cause the damage of the connections, including extreme temperatures, pervasive humidity or excessive loading, etc. With the development of structural health monitoring (SHM) technologies, machine learning algorithms and advanced sensor technologies, such as Piezoelectric Lead Zirconate Titanate (PZT), the health status of most large-scale structures can now be intelligently monitored in real time. In addition, image processing techniques can also be used in SHM to improve cost efficiency and convenience. Advanced SHM techniques for connections with PZT transducers and image processing technique are introduced in this dissertation. In one project, a PZT based SHM system for cuplock scaffolding connections was developed to monitor connection looseness. With further help of time reversal and wavelet packet analysis methods, different level of looseness for cuplock connections can be discerned. Another SHM technique developed in this dissertation was a wearable sensor device designed to be used for the real-time SHM of bolted flange joints. The wearable sensor device can clamp the sensors on the curved surface of the bolted flange joint without need for permanent adhesives. On top of sensing capabilities, this device provides vibration and electro-magnetic isolation as well as easy installation and disassembly. Thirdly, a data-driven SHM for the connection of space frame structure was developed based on wavelet packet analysis and machine learning algorithms. Using linear discriminant analysis and nonlinear support vector machine, the looseness degree of connection between the spherical connector and the supporting rod of the space frame structure can be predicted based on its frequency response signal. Finally, a vision-based SHM was described to detect the looseness of bolts in a flange. The detection method includes the perspective transformation of images taken from angled side views, identification of bolt positioning with digit recognition, and the smart detection of bolt rotation angles. The image processing method can help to increase efficiency and reduce cost in the looseness monitoring of bolted flange joints.