Material degradation identification in structural dynamics using stochastic finite elements

Chen, Qi ORCID: https://orcid.org/0000-0002-8804-0845 2024. Material degradation identification in structural dynamics using stochastic finite elements. PhD Thesis, Cardiff University. Item availability restricted.

Preview	PDF (Thesis) - Accepted Post-Print Version Download (60MB) | Preview
	PDF (Cardiff University Electronic Publication Form) Restricted to Repository staff only Download (181kB)

Abstract

Identifying structural system degradation is one of the primary objectives of structural health monitoring (SHM). This thesis addresses the inverse identification of material degradation in structures using vibration response data, accounting for material randomness and measurement noise. The discrepancy between the simulated and experimental responses, particularly the frequency response function (FRF), serves as an error metric to develop the degradation identification algorithm. Random field theory is employed to quantify spatially distributed material uncertainty. The stochastic finite element model update (SFEMU) method is applied to characterise material uncertainties and propagate them through the structural simulator. Unlike conventional interval-based uncertainty approaches, material variability is represented via probabilistic distributions, enabling a more rigorous quantification of uncertainty. The random field model provides a generic framework for inverse identification of material degradation in the presence of measurement noise and enhances the detection reliability. The structural material degradation (SMD) is treated as a local spatial material uncertainty. Rather than assigning a degradation factor to each element or quantify the damage using a series of parameters to locate the damage, the Karhunen-Loéve (KL) expansion method is implemented to spectrally decompose the material field into a finite dimensional random field representation, which can simulate any material property distributions with damage. Abstract iv The discrepancy between the analytical model and the KL-expansion-simulated model can be leveraged to assess simulation accuracy and facilitate the identification of SMD by minimizing this discrepancy using the SFEMU method. Following this, the differences between the simulated and experimental responses – specifically the numerical Frequency Response Function (FRF) – are employed as an error function to develop the identification algorithm and detect material degradation. Furthermore, the developed algorithm was applied in an experimental setup on a cantilever beam with different damage conditions to identify the SMD. Despite the experimental FRF being contaminated by measurement noise, which presents significant challenges for the inverse identification algorithm, the SMD was successfully identified.

Item Type:	Thesis (PhD)
Date Type:	Completion
Status:	Unpublished
Schools:	Schools > Engineering
Uncontrolled Keywords:	1. Hamiltonian Monte Carlo 2. Bayesian Inference 3. Random Field Simulation 4. Karhunen-Lo\'eve Expansion 5. Inverse Identification 6. Experimental FRF
Date of First Compliant Deposit:	3 July 2025
Last Modified:	03 Jul 2025 15:26
URI:	https://orca.cardiff.ac.uk/id/eprint/179525

Actions (repository staff only)

Edit Item