Structural Health Monitoring Optimization For Similar Structures Using A Hierarchical Bayesian Update

One of the most important issues in civil engineering is the detection of structural damages affecting the system performance at an early stage in order to prevent any catastrophic results. Two main approaches are adopted for structure monitoring: (i) periodical inspections and (ii) permanent monitoring. However, when relying on inspection results only, damages occurring between two consecutive inspections can not be detected which can be dangerous depending on the severity of the damage. Hence, for cost effective results, it is often recommended to permanently monitor structures using sensors in order to have continuous information about every single element of the structure. Since the implementation of such instruments is costly, one must optimize their configuration, number and location wise, in order to maximize the probability of detecting damages with a limited number of sensors. This configuration can be optimized, furthermore, when monitoring similar structures having at least one property in common. Thus, in this paper, we propose a methodology that optimizes the configuration of sensors for several similar structures by a genetic algorithm, and uses the obtained data in order to update, accordingly, the elements’ properties (i.e. The rigidity, the Young modulus) for each structure through a hierarchical Approximate Bayesian Computation (ABC). This methodology considers all uncertainties associated with the precision of the sensors results, the mechanical model and the degradation of the elements and takes advantage of the results obtained for any structure in order to update the information of all other similar structures which could save time, costs, and give more accurate results. A numerical application on two concrete frame structures is presented to illustrate the proposed methodology.