Bolted joints are key components in machines and infrastructure transferring loads and interconnecting parts. Loosening of bolted joints, particularly in terms of those utilized in aviation and railway engineering sector, can be vital and lead to catastrophic consequences. Monitoring of bolt loosening and furthermore quantitative determination of pre-loading force is therefore essential to provide alarming for in-time maintenance or replacement. This paper systematically describes an ultrasonic approach for real-time monitoring of bolt loosening based on electro-mechanical impedance (EMI) measured from a pre-designed PZT sensing network, which forms an electro-mechanical couple with the host structure. Finite element analysis of the tested couple is carried out to for both indication of signal features and validation purposes. Furthermore, a specifically well-tuned graph convolutional network (GCN) is utilized in bolt load determination, through learning the measured EMI signatures considering the sensor-sensor and sensor-bolt relationships in the PZT sensing network. The approach is expected to adaptively assess loading conditions of bolts on various types of host structures with a considerable confidence.
|