A physics-informed & data-driven ultrasonic guided wave structural health monitoring framework with uncertainty quantification

Gullapalli, Anirudh 2025. A physics-informed & data-driven ultrasonic guided wave structural health monitoring framework with uncertainty quantification. PhD Thesis, Cardiff University. Item availability restricted.

Preview	PDF - Accepted Post-Print Version Download (39MB) | Preview
	PDF (Cardiff University Electronic Publication Form) - Supplemental Material Restricted to Repository staff only Download (754kB)

Abstract

The drive towards predictive maintenance of safety–critical engineering structures, necessitates an on–board structural health monitoring system that can accurately as certain structural integrity. Current structural health monitoring (SHM) approaches for safety-critical systems face three fundamental limitations: (1) purely data-driven methods lack physical interpretability, while physics-based models struggle with real world complexity; (2) existing frameworks rarely quantify the uncertainties inherent in both measurements and model predictions; and (3) most systems require expensive instrumentation unsuitable for continuous on-board deployment. This thesis addresses these gaps through the conceptualization, development and experimental validation of a novel physics-informed and data driven structural health monitoring framework. The framework is composed of inexpensive edge-computing devices and a sparse piezo electric sensor array, operated by customized open-source Python scripts for continuous, periodic or event triggered ultrasonic signal acquisition. It also incorporates a convolutional neural network classification algorithm to distinguish essential signals from non-essential noise, ensuring only essential acoustic event representative signals undergo feature extraction. The value of the acquired data lies in extracting and mapping essential features to structural behaviour patterns altered by degradation. Towards this, a physics-informed approach was developed to calibrate the fundamental S0 and A0 parameters. These pa rameters not only generated highly accurate signal reconstructions at any propagation distance and angle but also captured progressive structural degradation due to cyclic compressive fatigue. A probabilistic Bayesian adaptive Metropolis Hastings–Markov Chain Monte Carlo approach quantified the complimentary aleatoric and epistemic uncertainties in experimental measurements and model estimates respectively. The resulting damage-sensitive modal parameters can identify damage types and assess severity, offering a foundation for a real-time, physics-informed structural health metric to monitor safety-critical engineering structures.

Item Type:	Thesis (PhD)
Date Type:	Completion
Status:	Unpublished
Schools:	Schools > Engineering
Uncontrolled Keywords:	1. Structural health monitoring 2. Ultrasonic Guided Waves 3. Digital Signal Processing 4. Bayesian Identification 5. Uncertainty Quantification 6. Digital Damage Identifiers
Date of First Compliant Deposit:	14 April 2026
Last Modified:	15 Apr 2026 08:55
URI:	https://orca.cardiff.ac.uk/id/eprint/186380

Actions (repository staff only)

Edit Item