Abstract

Use of a sensor network to provide adequate and reliable information is paramount for accurate damage detection of structures. However, unavoidably, deployed sensors are occasionally subject to failure faults, which, in turn, cause missing information. Placement of multiple backup sensors in a local region could overcome this difficulty and increase the sensor redundancy; however, this approach leads to a sensor clustering problem and higher costs in sensor deployment. Further, model uncertainty is another important issue that should be considered in a sensor network design. Accordingly, this work is dedicated to presenting a framework for optimization of sensor distribution that considers both sensor faults under uncertainty and sensor clustering for vibration-based damage detection. Based on the effective independence method, the first design objective is newly formulated to consider sensor faults under uncertainty. Moreover, a novel index that is universally applicable for any type of structure is proposed to evaluate sensor clustering, which is treated as the second objective. The non-dominated sorting genetic algorithm II is adopted to solve this multi-objective optimization problem, and Monte Carlo simulation (MCS) is employed for uncertainty analysis in the first objective. To reduce computation costs, real performance evaluations in MCS are replaced with Gaussian process regression models. Based on the vibration information achieved from optimized sensors, an optimization-based damage detection process is applied to validate the optimal sensor layout. Three case studies (i.e., a cantilever beam, a laminated composite structure, and a spatial frame) are presented to demonstrate the effectiveness and applicability of the developed framework.