Vi-Detect: Fine-Grained Vibration-Based Component Looseness Detection Using Smartphones

In civil engineering, many structures, like steel truss bridges, warehouses, and wood structures, are interconnected using bolts. However, bolts can loose under shock or vibration, which may lead to overall structural failure or even serious accidents. Early detection of component defects is crucial for the safety of the structure. In this research, we propose Vi-Detect, a system that uses smartphones’ built-in accelerometers to detect component looseness in structures. In contrast to existing vibration-based detection methods that provide only binary results (tight or loose), the major novelty of Vi-Detect lies in that it can detect the continuous extent/degree to which a connection has been loosened. To achieve this, Vi-Detect measures a structure’s motion response to an external vibration excitation, and then reverse-engineers a second-order mass-spring-damper model for the structure, whose theoretical motion response to the same external excitation matches with the measured response. The value of key parameters in this model, such as the stiffness (resists deformation) and damping (regulate vibration dissipation) coefficients, are then used to calculate the continuous degree of looseness (a.k.a. tightness index) of the connections. Extensive experiments on real-world testbed with wood and steel components have been conducted to validate Vi-Detect’s effectiveness and accuracy.