Abstract:Numerical simulation is powerful to study nonlinear solid mechanics problems. However, mesh-based or particle-based numerical methods suffer from the common shortcoming of being time-consuming, particularly for complex problems with real-time analysis requirements. This study presents a clustering adaptive Gaussian process regression (CAG) method aiming for real-time prediction for nonlinear structural responses in solid mechanics. It is a data-driven machine learning method featuring a small sample size, high accuracy, and high efficiency, leveraging nonlinear structural response patterns. Similar to the traditional Gaussian process regression (GPR) method, it operates in offline and online stages. In the offline stage, an adaptive sample generation technique is introduced to cluster datasets into distinct patterns for demand-driven sample allocation. This ensures comprehensive coverage of the critical samples for the solution space of interest. In the online stage, following the divide-and-conquer strategy, a pre-prediction classification categorizes problems into predefined patterns sequentially predicted by the trained multi-pattern Gaussian process regressor. In addition, dimension reduction and restoration techniques are employed in the proposed method to enhance its efficiency. A set of problems involving material, geometric, and boundary condition nonlinearities is presented to demonstrate the CAG method's abilities. The proposed method can offer predictions within a second and attain high precision with only about 20 samples within the context of this study, outperforming the traditional GPR using uniformly distributed samples for error reductions ranging from 1 to 3 orders of magnitude. The CAG method is expected to offer a powerful tool for real-time prediction of nonlinear solid mechanical problems and shed light on the complex nonlinear structural response pattern.

What problem does this paper attempt to address?

This paper attempts to address the problem of achieving real-time prediction in nonlinear solid mechanics. Specifically, existing grid-based or particle-based numerical methods are time-consuming when dealing with complex problems, especially in scenarios requiring real-time analysis. To overcome this challenge, the paper proposes a Clustering Adaptive Gaussian Process Regression (CAG) method, aiming to achieve real-time prediction of nonlinear solid mechanics problems through nonlinear structural response patterns, leveraging the characteristics of a small number of samples, high accuracy, and high efficiency. ### Background and Problem Statement 1. **Background**: - Nonlinear solid mechanics problems involve complex material, geometric, and boundary condition nonlinearities, typically requiring numerical simulation algorithms for study. - Grid-based or particle-based numerical methods are very time-consuming when dealing with large-scale problems, particularly in scenarios requiring immediate evaluation, such as high-speed impact, seismic loads, etc. - Data-driven methods offer a promising path that can significantly improve prediction speed and are considered the new scientific paradigm following computational science. 2. **Limitations of Existing Methods**: - Large-sample data-driven methods require a vast amount of data for model training to ensure prediction accuracy and generalization ability, but these methods face issues of interpretability and data availability in complex nonlinear solid mechanics problems. - Small-sample data-driven methods can achieve the required performance on limited datasets, avoiding the time-consuming and expensive data preparation process, but selecting appropriate training samples and avoiding interference from low-relevance samples is a challenge. ### Proposed Method The paper proposes a Clustering Adaptive Gaussian Process Regression (CAG) method, which combines demand-driven sample allocation in the offline phase and a divide-and-conquer strategy in the online phase, including the following steps: 1. **Offline Phase**: - **Adaptive Sample Generation**: Classify samples into different patterns using unsupervised learning techniques and generate new training samples to ensure comprehensive coverage of all patterns. - **Multi-Pattern Gaussian Process Regressor Training**: Perform supervised learning on the classified training samples for each pattern to form multi-pattern Gaussian process regressors, and further improve efficiency using dimensionality reduction techniques. 2. **Online Phase**: - **Pre-Prediction Classification**: Classify the given case using predefined patterns. - **Prediction**: Use the trained multi-pattern Gaussian process regressors to predict based on the classified pattern, reducing interference from irrelevant training samples. - **Physical Field Reconstruction**: Reconstruct meaningful physical fields from the predicted reduced-dimension variables to represent structural behavior. ### Experimental Validation The paper validates the accuracy and efficiency of the CAG method in nonlinear solid mechanics problems through a series of numerical cases, including complex nonlinear problems of hyperelastic cylinders under axial tension and compression, elastoplastic behavior of T-shaped structures under extreme torsion, high-speed impact on rings, and dynamic crack branching in brittle plates. ### Conclusion The CAG method can complete predictions within seconds and achieve high accuracy with only about 20 samples, improving error reduction by 1 to 3 orders of magnitude compared to traditional Gaussian process regression methods with uniformly distributed samples. This method is expected to become a powerful tool for real-time prediction of nonlinear solid mechanics problems and provides new insights for studying complex nonlinear structural response patterns.

A clustering adaptive Gaussian process regression method: response patterns based real-time prediction for nonlinear solid mechanics problems

An intelligent response surface method for analyzing slope reliability based on Gaussian process regression

Deformation prediction of foundation pit using Gaussian Process machine learning

Fast analysis of structural reliability using gaussian process classification based dynamic response surface method

Nonlinear response modelling of material systems using constrained Gaussian processes

Data-driven spatiotemporal modeling for structural dynamics on irregular domains by stochastic dependency neural estimation

A Gaussian Process-Based Dynamic Surrogate Model for Complex Engineering Structural Reliability Analysis

Gaussian Process Emulation For Big Data In Data-Driven Metamaterials Design

Gaussian Process-Based Dynamic Response Surface Method For Estimating Slope Failure Probability

FEM-Cluster Based Reduction Method for Efficient Numerical Prediction of Effective Properties of Heterogeneous Material in Nonlinear Range

A Nested Non-Intrusive Stochastic Isogeometric Method for Nonlinear Thermal Vibration of Fgm Plates under Random Loading

Fast Parameter Optimization for High-Fidelity Crystal Plasticity Simulation Using Active Learning

Probabilistic parameter estimation and reliability assessment of nonlinear structures based on generative adversarial networks with Gaussian process regression

Probabilistic analysis of tunnel convergence in spatially variable soil based on Gaussian process regression

A novel sparse Gaussian process regression with time-aware spatiotemporal kernel for remaining useful life prediction and uncertainty quantification of bearings

Gaussian process regression based remaining fatigue life prediction for metallic materials under two-step loading

Mechanistically Informed Data-Driven Modeling of Cyclic Plasticity Via Artificial Neural Networks

Clustering discretization methods for generation of material performance databases in machine learning and design optimization

Learning nonlinear constitutive models in finite strain electromechanics with Gaussian process predictors

Probabilistic Prediction of Joint Shear Strength Using Gaussian Process Regression with Anisotropic Compound Kernel

WearGP: A computationally efficient machine learning framework for local erosive wear predictions via nodal Gaussian processes