Population-based Structural Health Monitoring: the Network of Structures

Structural Health Monitoring (SHM), at its core, is the process of monitoring a system or structure with the objective of utilising the acquired data to access the overall condition - the health - of the object in question. One problem of such an approach is acquiring enough data to specify the `health' state of a structure. Population-based Structural Health Monitoring (PBSHM) is a recent development within the SHM community which attempts to bypass the data bottlenecks present in the `classic' SHM scenario.

The aim of PBSHM is that by monitoring multiple structures - the population - one can gain additional insights into the health of a particular structure when using population data, compared to the insights available when using only a single structure's data. PBSHM operates under the premise that learnt knowledge may be shared across structures in the population; however, before any knowledge is shared, a similarity between structures is first established, to guide if an attempt to share knowledge should occur. The work conducted in this thesis is focussed on the similarity-assessment portion of PBSHM and is divided into two parts.

The first part of the thesis, focusses on developing the required language and ecosystem for determining the similarity of structures. To achieve this, a radical reconstruction of the core concepts, definitions, and language of an Irreducible Element (IE) model - the vehicle used within PBSHM to describe a structure - was required, to provide a standardised representation of structures. A novel ecosystem is then introduced to provide a shared-domain context for PBSHM similarity computations: network, framework, and database.

The second part of the thesis, focusses on the problems faced during the similarity assessment process. When an IE model is curated, there is inherent author bias present within the model which introduces nuanced variations within the model. The Canonical Form is introduced as the potential vehicle for facilitating the network to adapt to these aforementioned variations. A case study is further given of generating an IE model for a real-world aircraft, a new graph notation is also introduced to enable concise pictorial representations of IE models.