Abstract:The physics-driven method via finite element model and the data-driven methods via supervised learning is commonly used in the analysis of structural damage identification. The finite element model can be susceptible to environmental noise and modeling errors, which leads to model identification results deviating from the real damage. The supervised learning method faces the challenge of insufficient labeled damage data, which limits the learned model in identifying damage scenarios that are not contained in the training data. In this study, a novel physics-guided data-driven method is proposed to incorporate physical perception into the data-driven method for identifying structural damage. Specifically, the physics from the finite element model (e.g., structural dynamic characteristics) is incorporated into the neural network model to guide the damage feature learning from scarce measured data. In addition, a loss function with physical significance is developed to evaluate the discrepancy between the finite element model and the measured data, which guides the neural network to learn the damage features of the real structure. The learned neural network model has the potential to improve the damage detection results. The proposed methodology is validated by a numerical case study of a suspension bridge model and an experimental study of a frame structure.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to combine the physical knowledge of the finite element model with the advantages of supervised learning methods in structural damage identification and overcome the limitations of a single method. Specifically: 1. **Limitations of the finite element model**: - The finite element model is easily affected by environmental noise and modeling errors, resulting in the model identification results deviating from the actual damage situation. - Modeling errors (such as stiffness errors, material density errors, etc.) and environmental noise are difficult to avoid in practical applications, and these factors may lead to inaccurate damage diagnosis results of the finite element model. 2. **Limitations of supervised learning methods**: - The main challenge faced by supervised learning methods is the lack of sufficient labeled damage data, which limits the model's ability to identify damage scenarios not included in the training data. - For large - scale structures (such as long - span bridges and high - rise frame buildings), generating damage labels is very time - consuming and costly. 3. **Proposed method**: - The paper proposes a Physics - Guided Deep Neural Network (PGDNN) method, aiming to incorporate physical awareness into data - driven methods to improve the accuracy of structural damage identification. - Specifically, by incorporating the physical characteristics (such as structural dynamic characteristics) in the finite element model into the neural network model, it guides the learning of damage features from sparse measurement data. - A loss function with physical significance has been developed to evaluate the difference between the finite element model and the measurement data, thereby guiding the neural network to learn the damage features of the real structure. 4. **Objectives**: - Improve the accuracy of structural damage detection, especially in the presence of modeling errors and environmental noise. - By combining the advantages of physical models and data - driven methods, reduce the dependence on a large amount of labeled data and improve the generalization ability of the model. In summary, the core problem of this paper is to solve the deficiencies of existing physics - driven and data - driven methods in structural damage identification by proposing a new physics - guided deep neural network method, and improve the accuracy and robustness of damage identification.

Physics-guided deep neural network for structural damage identification

Model-informed Deep Learning Strategy with Vision Measurement for Damage Identification of Truss Structures

Structural damage identification via physics-guided machine learning: a methodology integrating pattern recognition with finite element model updating

Bridge damage identification under the moving vehicle loads based on the method of physics-guided deep neural networks

Deep learning-based bridge damage identification approach inspired by internal force redistribution effects

Model-based Damage Detection through Physics Guided Learning

Structural damage identification based on autoencoder neural networks and deep learning

A damage-informed neural network framework for structural damage identification

A Novel Structural Damage Identification Method Using a Hybrid Deep Learning Framework

A Time–Frequency-Based Data-Driven Approach for Structural Damage Identification and Its Application to a Cable-Stayed Bridge Specimen

A Method of Structural Damage Identification Using Neural Networks Based on Static Displacements and Natural Frequencies

Physics-Informed Neural Networks for Missing Physics Estimation in Cumulative Damage Models: A Case Study in Corrosion Fatigue

Structural Damage Detection with Automatic Feature‐Extraction Through Deep Learning

Unsupervised structural damage identification based on covariance matrix and deep clustering

Damage identification of steel bridge based on data augmentation and adaptive optimization neural network

Vibration-based damage detection for bridges by deep convolutional denoising autoencoder

Damage identification based on the inner product matrix and parallel convolution neural network for frame structure

Unsupervised Quantitative Structural Damage Identification Method Based on BiLSTM Networks and Probability Distribution Model

Application of Graph Convolutional Neural Networks Combined with Single-Model Decision-Making Fusion Neural Networks in Structural Damage Recognition

Deep Learning Based Structural Damage Identification for the Strain Field of a Subway Bolster

Limited-Data-Driven Machine Learning in Structural Health Diagnosis