Sensor distributions for structural monitoring: a correlation study

Deciding on the spatial distribution of output sensors for vibration-based structural health monitoring (SHM) is a task that has been, and still is, studied extensively. Yet, when referring to the conventional damage characterization hierarchy, composed of detection, localization, and quantification, it is primarily the first component that has been addressed with regard to optimal sensor placement. In this particular context, a common approach is to distribute sensors, of which the amount is determined a priori, such that some scalar function of the probability of detection for a pre-defined set of damage patterns is maximized. Obviously, the optimal sensor distribution, in terms of damage detection, is algorithm-dependent, but studies have showed how correlation generally exists between the different strategies. However, it still remains a question how this “optimality” correlates with the optimal sensor distribution for the next component in the damage characterization triad, namely, localization. Realizing that interrogating for the location of damage is only of relevance if deterioration has already been detected, it is paramount that a sensor distribution selected for damage localization also provides a favorable damage detection resolution. In the present paper, the correlation, or lack of such, in optimal sensor distributions is investigated for well-established damage detection schemes and a recently proposed localization method. This is done in the context of simulations with a simple chain-like system, which is instrumented with a fixed number of sensors and examined for particular damage patterns. By ranking all the possible sensor distributions according to, respectively, probability of detection and probability of localization, a sub-optimal distribution is found that maximizes the sum of these probabilities.