Creating a network of structures based on physical similarity

Effective structural health monitoring (SHM) requires large amounts of data representing the normal condition of a structure as well as any damage conditions. However, it is not always feasible to obtain these data; for example, it is not economical to obtain damage-state data for a new bridge. To address this problem, a new framework is being explored called population-based structural health monitoring (PBSHM), which proposes that if two structures are sufficiently similar, then data can be shared between them. Tools which enable the sharing of data, such as the transfer of models and damage classifiers, have been explored in previous work; as have methods for assessing the similarity of structures. This paper will describe how it may be possible to link structures based on their physical similarity in such a way that creates a network, with communitities of similar structures. Within these communities, data can be shared between structures. Forming these communities in a way that is computationally efficient while still avoiding missing possible links is not straightforward. This paper outlines some of the considerations that must be taken into account for solving this problem.