Real-Time Structural Health Monitoring with Bayesian Neural Networks: Distinguishing Aleatoric and Epistemic Uncertainty for Digital Twin Frameworks

2025-12-02

AI Summary

Real-time, spatially resolved uncertainty quantification remains challenging in structural health monitoring (SHM). Method: This paper proposes a full-field uncertainty modeling framework integrating principal component analysis (PCA), Bayesian neural networks (BNNs), and Hamiltonian Monte Carlo (HMC) inference. Leveraging only sparse strain sensor measurements, the framework reconstructs high-fidelity full-field strain distributions and—crucially—disentangles aleatoric uncertainty (from measurement noise) and epistemic uncertainty (from model inadequacy) at the pixel level. Contribution/Results: The method achieves high reconstruction accuracy (R² > 0.9 for carbon fiber specimens) and enables interpretable, spatially localized confidence assessment. It demonstrates robustness under crack-induced strain singularities and supports trustworthy, real-time damage diagnosis and decision-making within digital twin applications.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Relational Probabilistic Models
• Intelligent Robots: State Estimation

Application Category

• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs
• Web Mining and Content Analysis: Web data provenance, reliability, and authenticity
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Reliable real-time analysis of sensor data is essential for structural health monitoring (SHM) of high-value assets, yet a major challenge is to obtain spatially resolved full-field aleatoric and epistemic uncertainties for trustworthy decision-making. We present an integrated SHM framework that combines principal component analysis (PCA), a Bayesian neural network (BNN), and Hamiltonian Monte Carlo (HMC) inference, mapping sparse strain gauge measurements onto leading PCA modes to reconstruct full-field strain distributions with uncertainty quantification. The framework was validated through cyclic four-point bending tests on carbon fiber reinforced polymer (CFRP) specimens with varying crack lengths, achieving accurate strain field reconstruction (R squared value>0.9) while simultaneously producing real-time uncertainty fields. A key contribution is that the BNN yields robust full-field strain reconstructions from noisy experimental data with crack-induced strain singularities, while also providing explicit representations of two complementary uncertainty fields. Considered jointly in full-field form, the aleatoric and epistemic uncertainty fields make it possible to diagnose at a local level, whether low-confidence regions are driven by data-inherent issues or by model-related limitations, thereby supporting reliable decision-making. Collectively, the results demonstrate that the proposed framework advances SHM toward trustworthy digital twin deployment and risk-aware structural diagnostics.

Problem

• Real-time monitoring of structural health with uncertainty quantification
• Distinguishing data noise from model limitations in strain field reconstruction
• Supporting reliable decision-making for digital twin frameworks in SHM

Innovation

• Bayesian neural network for full-field strain reconstruction
• Principal component analysis mapping sparse gauge data
• Hamiltonian Monte Carlo inference for uncertainty quantification