Modal Decomposition and Identification for a Population of Structures Using Physics-Informed Graph Neural Networks and Transformers

2025-05-06

AI Summary

This work addresses unsupervised joint identification of modal parameters—natural frequencies, damping ratios, and mode shapes—in population-scale structural health monitoring under sparse dynamic measurements, particularly in the presence of varying external excitations and structural configurations. We propose the first purely physics-driven, label-free end-to-end framework that integrates graph neural networks (GNNs) to encode structural topology, Transformers to model long-range temporal dependencies, and modal decomposition theory as hard physical constraints, enforced via a physics-informed loss function. Evaluated on both numerical simulations and experimental data, the method successfully decouples single-degree-of-freedom modal responses from sparse vibration measurements and achieves high-accuracy identification of all three modal attributes. It significantly outperforms conventional modal analysis techniques and, for the first time, enables robust, excitation- and configuration-invariant modal identification across populations of structures.

Technology Category

• Machine Learning: Multimodal Learning
• Computer Vision: Multi-modal Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Web Mining and Content Analysis: Bridging structured and unstructured data

Abstract

Modal identification is crucial for structural health monitoring and structural control, providing critical insights into structural dynamics and performance. This study presents a novel deep learning framework that integrates graph neural networks (GNNs), transformers, and a physics-informed loss function to achieve modal decomposition and identification across a population of structures. The transformer module decomposes multi-degrees-of-freedom (MDOF) structural dynamic measurements into single-degree-of-freedom (SDOF) modal responses, facilitating the identification of natural frequencies and damping ratios. Concurrently, the GNN captures the structural configurations and identifies mode shapes corresponding to the decomposed SDOF modal responses. The proposed model is trained in a purely physics-informed and unsupervised manner, leveraging modal decomposition theory and the independence of structural modes to guide learning without the need for labeled data. Validation through numerical simulations and laboratory experiments demonstrates its effectiveness in accurately decomposing dynamic responses and identifying modal properties from sparse structural dynamic measurements, regardless of variations in external loads or structural configurations. Comparative analyses against established modal identification techniques and model variations further underscore its superior performance, positioning it as a favorable approach for population-based structural health monitoring.

Problem

• Identifying structural modes from sparse dynamic measurements
• Decomposing MDOF responses into SDOF modal components
• Enabling unsupervised modal analysis across diverse structures

Innovation

• Integrates GNNs and transformers for modal decomposition
• Uses physics-informed loss for unsupervised learning
• Decomposes MDOF to SDOF for accurate modal identification