Towards Population-Based Structural Health Monitoring, Part III: Graphs, Networks and Communities

Abstract

Population-based structural health monitoring opens up the possibility of using information from a population of structures to provide extra information for each individual structure. For example, population-based structural health monitoring could provide improved damage-detection within a homogeneous population of structures by defining a normal condition across a population of structures, which was robust to environmental variation. Furthermore, in cases where structures are sufficiently similar, damage location, assessment, and classification labels could be transferred, increasing the damage labels available for each structure. To determine whether two structures are sufficiently similar requires the comparison of some representation of the structure. In fields such as bioinformatics and computer science, attributed graphs are often used to determine structural similarity. This paper will describe methods for comparing the topology attributes of two such graphs. The algorithm described is suited to population-based structural health monitoring as it provides matches between two graphs which have physical significance. This paper will also describe the process of comparing hierarchical attributes to determine the level of knowledge transfer possible between two structures.