Structural health monitoring of composite structures: low-velocity impact localisation and force reconstruction
File(s)
Author(s)
Xiao, Dong
Type
Thesis
Abstract
This thesis addresses key challenges in structural health monitoring (SHM) of composite structures using passive sensing, with a particular focus on solving the inverse problem of impact identification. It develops a comprehensive framework for robust, efficient, interpretable, and scalable methodologies for impact localisation and force reconstruction, integrating structural dynamics analysis, guided wave propagation, advanced signal processing, and machine learning.
To enhance impact localisation, several novel approaches are introduced. First, a data-driven method leveraging Gaussian Process Regression with composite kernel design and Bayesian model fusion is developed, ensuring high accuracy and robustness across varying environmental and operational conditions. Second, a hybrid physics-data approach combines wave propagation physics with data-driven techniques, significantly improving generalisability in complex composite structures. Third, a general Bayesian probabilistic framework for impact localisation is proposed, adaptable to structural complexity and available data. These methods are underpinned by rigorous theoretical analyses of wave propagation in composite plates using plate theories.
For impact force reconstruction, an adaptive wavelet-based regularisation technique is introduced to mitigate its ill-posed nature, enabling efficient and accurate force identification when the structural transfer functions are known. Furthermore, to achieve precise force estimation at any impact location using sparse impact data, an physics-guided impact identification method is developed. This approach integrates sparse data-driven physics modelling with data-augmented machine learning, enhancing interpretability and accuracy across diverse impact scenarios.
The proposed methodologies are rigorously validated through extensive experiments on representative composite structures. Results demonstrate their effectiveness, robustness, and scalability, marking a significant advancement in SHM for composite aerostructures. This research contributes to improving the reliability, safety, and sustainability of next-generation aerospace systems.
To enhance impact localisation, several novel approaches are introduced. First, a data-driven method leveraging Gaussian Process Regression with composite kernel design and Bayesian model fusion is developed, ensuring high accuracy and robustness across varying environmental and operational conditions. Second, a hybrid physics-data approach combines wave propagation physics with data-driven techniques, significantly improving generalisability in complex composite structures. Third, a general Bayesian probabilistic framework for impact localisation is proposed, adaptable to structural complexity and available data. These methods are underpinned by rigorous theoretical analyses of wave propagation in composite plates using plate theories.
For impact force reconstruction, an adaptive wavelet-based regularisation technique is introduced to mitigate its ill-posed nature, enabling efficient and accurate force identification when the structural transfer functions are known. Furthermore, to achieve precise force estimation at any impact location using sparse impact data, an physics-guided impact identification method is developed. This approach integrates sparse data-driven physics modelling with data-augmented machine learning, enhancing interpretability and accuracy across diverse impact scenarios.
The proposed methodologies are rigorously validated through extensive experiments on representative composite structures. Results demonstrate their effectiveness, robustness, and scalability, marking a significant advancement in SHM for composite aerostructures. This research contributes to improving the reliability, safety, and sustainability of next-generation aerospace systems.
Version
Open Access
Date Issued
2025-06-03
Date Awarded
2025-09-01
Copyright Statement
Attribution-NonCommercial 4.0 International Licence (CC BY-NC)
License URL
Advisor
Sharif-Khodaei, Zahra
Aliabadi, M. H.
Sponsor
Imperial College London
China Scholarship Council
Grant Number
[2021]339
Publisher Department
Department of Aeronautics
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)