Design of a monitoring system for a long-span suspension bridge: optimal sensor placement

Abstract

The design of systems for structural health monitoring is an exercise that requires careful planning of optimal sensor locations. The number of possible locations usually outnumbers the number of available sensors, which means that solving the problem of optimal sensor placement (OSP) is necessary for a cost-effective network. This paper addresses the design of OSP for a new monitoring system on a long-span suspension bridge. The foremost interest is the response effect of wind loading on the bridge. A finite element model of the bridge is created based on blueprint drawings, and the system model consists of 18 target modes (with frequencies in the range [0.05, 0.6] Hz). 22 triaxial force-balance MEMS accelerometers are planned to be deployed. In the design, the sensors are designated into local groups consisting of two or four accelerometers, which are wired to the same data acquisition unit. Two optimization approaches are tried: backward sequential sensor placement and genetic algorithms. The objective metric is the maximization of the determinant of the Fisher information matrix. Six different types of models for the prediction error covariance are tested; one with no correlation and five with varying degrees of error correlation, which is dictated by the sensor distance and a characteristic correlation length. The resulting configurations are compared and discussed.