Aims and objectives
Analogous to the human nervous system, the Structural Health Monitoring (SHM) concept aims to monitor a structural system and assess its current and future performance. However, the SHM concept is subjected to numerous sources of uncertainty, both during the monitoring phase (noise, discretisation errors, selection of appropriate frequency bandwidth) or the data interpretation phase (modelling and interpolation errors, unknown parameters, separation of relevant features from other confounding influences). The current PhD project proposed a hybrid modular Bayesian uncertainty assessment for improvement of SHM. The framework is hybrid, in the sense that it uses a physics-based model, and Gaussian processes (mrGp) which are trained against data, for uncertainty quantification. The mrGp act as emulators of the model response surface and its model discrepancy, also quantifying observation error, parametric and interpolation uncertainty. Two extensions of the original framework have been implemented, specifically for structural identification and measurement system design, with applications to simulated, small-scale and full-scale structures.
Methodologies
Remarkably, the aforementioned framework can be extended for structural identification, measurement system design or damage detection. The ability to learn from data and simulations of a physics-based model at the same time, is the most novel aspect which enhances each of these SHM topics. Some of the flowcharts of these methodologies can be seen in the below links.
Case-studies
