Voltage-Based Unsupervised Learning Framework for Bridge Damage Detection in Simultaneous Energy Harvesting and Sensing Systems

In this study, piezoelectric energy harvesters (PEHs) are designed to offer dual functionality in structural health monitoring (SHM): harvesting electric power from bridge vibrations while serving as intrinsic damage sensors. This strategy utilises the voltage signal directly as the sensing input, eliminating the need for traditional sensing modules and thereby reducing system complexity and energy consumption. A bi-objective optimisation framework is proposed to maximise both power output and damage detection accuracy of a PEH modelled as a composite cantilevered Kirchhoff-Love plate. Voltage responses under realistic bridge inputs are predicted via isogeometric analysis. The approach is validated in two scenarios: a numerical vehicle-bridge interaction model and a laboratory-scale beam test using a toy car, each evaluated in both healthy and damaged states. Unsupervised damage detection is achieved using a convolutional variational autoencoder (CVAE) trained solely on healthy voltage signatures. The NSGA-II algorithm is applied to explore trade-offs between energy yield and sensing precision, including parametric studies on damage severity, damage location, and harvester geometry. Results indicate that optimised PEHs not only act as an effective filter and sensing component but also outperform traditional acceleration-based sensing, improving damage detection accuracy by 13% while reducing energy consumption by 98%. The multi-parameter design space further highlights the importance of bi-objective optimisation due to variations in performance even under resonant conditions. These findings demonstrate the feasibility of replacing traditional sensors with lightweight, self-powered PEHs and pave the way for sustainable simultaneous energy harvesting and sensing (SEHS) systems.