The Effects of an Extended Sensitivity Analysis of Sensor Configurations for Bridge Damage Detection Using Experimental Data

The damage detection capabilities of sensor setups are essential for any structural health monitoring (SHM) system. In this chapter, the performance of different subsets of sensor configurations selected from a set of 40 accelerometers is evaluated using metrics such as misclassification rate, false positive (FP), and false negative (FN) indications of damage. The subsets of sensor configurations are based on experimental data from a benchmark study that involved capturing the dynamic behavior of a full-scale steel bridge in undamaged and damaged conditions of the bridge. Several iterations with new subsets of decreasing size are generated by the elimination of random sensors. These subsets are then tested using Mahalanobis squared distance (MSD) as the novelty detection algorithm. Additionally, a manual selection of subsets is evaluated, where the sensors located farthest from the damages are eliminated. The results highlight the advantages of a dense sensor network and indicate a complex mechanism behind the damage detection capabilities of sensor networks with a clear trend of inverse proportionality between the sensor set size and FN indications of damage.