Abstract:Finite element modeling is a well-established tool for structural analysis, yet modeling complex structures often requires extensive pre-processing, significant analysis effort, and considerable time. This study addresses this challenge by introducing an innovative method for real-time prediction of structural static responses using DeepOnet which relies on a novel approach to physics-informed networks driven by structural balance laws. This approach offers the flexibility to accurately predict responses under various load classes and magnitudes. The trained DeepONet can generate solutions for the entire domain, within a fraction of a second. This capability effectively eliminates the need for extensive remodeling and analysis typically required for each new case in FE modeling. We apply the proposed method to two structures: a simple 2D beam structure and a comprehensive 3D model of a real bridge. To predict multiple variables with DeepONet, we utilize two strategies: a split branch/trunk and multiple DeepONets combined into a single DeepONet. In addition to data-driven training, we introduce a novel physics-informed training approaches. This method leverages structural stiffness matrices to enforce fundamental equilibrium and energy conservation principles, resulting in two novel physics-informed loss functions: energy conservation and static equilibrium using the Schur complement. We use various combinations of loss functions to achieve an error rate of less than 5% with significantly reduced training time. This study shows that DeepONet, enhanced with hybrid loss functions, can accurately and efficiently predict displacements and rotations at each mesh point, with reduced training time.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to use deep - learning techniques, especially the DeepONet - based method, to efficiently predict the static responses of structures under different load conditions, thereby reducing the large amount of pre - processing, analysis time and computational resources required by the traditional finite element method (FEM) when modeling complex structures. Specifically, by introducing a physical - information - based loss function based on structural stiffness, the paper improves the accuracy and efficiency of DeepONet in predicting structural responses while reducing the training time. This method can generate solutions for the entire structural domain within a few seconds without the need for extensive re - modeling and analysis for each new case. ### Core problems of the paper 1. **Improving prediction efficiency**: The traditional finite element method requires a large amount of time and computational resources when dealing with complex structures. The method proposed in the paper aims to achieve fast and accurate structural response prediction through machine - learning techniques, especially DeepONet. 2. **Ensuring physical consistency**: To ensure that the prediction results conform to physical laws, the paper introduces a physical - information - based loss function based on structural stiffness, including the principles of energy conservation and static equilibrium. These loss functions ensure that the model's predictions not only rely on data but also conform to basic physical laws. 3. **Multivariable prediction**: The paper explores how to use DeepONet to simultaneously predict the responses of multiple degrees of freedom, such as displacement and rotation, which is very important for practical engineering applications. ### Main contributions 1. **Innovative DeepONet architecture**: The paper proposes a multi - output strategy based on DeepONet, including independent DeepONet and split - branch / trunk networks, to handle multiple output variables. 2. **Physical - information - based loss functions**: New physical - information - based loss functions based on energy conservation and static equilibrium are developed. These loss functions utilize pre - calculated stiffness matrices and avoid the use of complex partial differential equations (PDEs). 3. **Efficient training method**: By introducing the Schur complement method, the dimension of the problem is reduced, thereby significantly reducing the training time and computational cost. 4. **Verification in practical applications**: The paper verifies the effectiveness of the proposed method on two structures, including a simple 2D beam structure and a real - world 3D bridge model (KW - 51). ### Mathematical formulas - **Energy conservation loss function**: \[ L_{\text{EC}}=\frac{1}{N}\sum_{i = 1}^{N}\left\|G(u_i)(a)^T\cdot K\cdot G(u_i)(a)-G(u_i)(a)^T\cdot F\right\| \] where \(G(u_i)(a)\) is the output function value predicted by DeepONet, \(K\) is the global stiffness matrix, and \(F\) is the global force. - **Static equilibrium loss function**: \[ L_{\text{SE - S}}=\frac{1}{N}\sum_{i = 1}^{N}\left\|S_i\cdot G(u_i)(a)-F_{c,i}\right\| \] where \(S_i\) is the Schur complement matrix and \(F_{c,i}\) is the simplified global force. - **Combined loss function**: \[ L_{\text{DD + EC}}=\frac{1}{N}\sum_{i = 1}^{N}\left(\frac{\left\|G_{\text{True}}(u_i)(a)-G(u_i)(a)\right\|}{\left\|G_{\text{True}}(u_i)(a)\right\|}+\left\|G(u_i)(a)^T\cdot K\cdot G(u_i)(a)-G(u_i)(a)^T\cdot F\right\|\right) \] \[ L_{\text{DD + SE - S}}=\frac{1}{N}\sum_{i = 1}^{N}\left(

Physics-informed DeepONet with stiffness-based loss functions for structural response prediction

Time Series Estimation Based on Deep Learning for Structural Dynamic Nonlinear Prediction

Structural Surrogate Model and Dynamic Response Prediction with Consideration of Temporal and Spatial Evolution: an Encoder–Decoder ConvLSTM Network

Novel DeepONet architecture to predict stresses in elastoplastic structures with variable complex geometries and loads

DeepPhysics: a physics aware deep learning framework for real-time simulation

Material-Response-Informed DeepONet and its Application to Polycrystal Stress-strain Prediction in Crystal Plasticity

Displacement Reconstruction Based on Physics-Informed DeepONet Regularizing Geometric Differential Equations of Beam or Plate

Material-Response-Informed DeepONet and Its Application to Polycrystal Stress–Strain Prediction in Crystal Plasticity

Advanced Deep Operator Networks to Predict Multiphysics Solution Fields in Materials Processing and Additive Manufacturing

Deep learning operator network for plastic deformation with variable loads and material properties

Combination of physics‐based and data‐driven modeling for nonlinear structural seismic response prediction through deep residual learning

Physics informed neural network for dynamic stress prediction

Deep neural operator for learning transient response of interpenetrating phase composites subject to dynamic loading

An Advanced Physics-Informed Neural Operator for Comprehensive Design Optimization of Highly-Nonlinear Systems: An Aerospace Composites Processing Case Study

Neuro-DynaStress: Predicting Dynamic Stress Distributions in Structural Components

Novel Physics-Informed Neural Network Approach for Dynamic and Static Displacement Reconstruction Via Strain and Acceleration

Physics-guided deep neural network for structural damage identification

Time-Continuous Energy-Conservation Neural Network for Structural Dynamics Analysis

Sequential Deep Operator Networks (S-DeepONet) for predicting full-field solutions under time-dependent loads

FEA-Net: A physics-guided data-driven model for efficient mechanical response prediction