Abstract:Accurately obtaining physics model information is essential for comprehending the mechanisms of physical dynamics. However, the inherent complexity of these models, especially in scenarios involving turbulence and irregular terrains, often requires significant computational resources to acquire high-fidelity (HF) data. While deep learning-based approaches have demonstrated considerable potential, they generally necessitate large volumes of training data and often face challenges in efficiently integrating physical information, such as derivatives. Leveraging the concept of multi-fidelity (MF) surrogate model and Gaussian processes (GPs), we introduce a novel approach, fast multi-fidelity Gaussian processes with derivatives (FMGPD), in which the posterior mean of the low-fidelity (LF) model replaces the prior mean of the HF model, with the relationship between the two encapsulated by an unknown function. To speed up convergence during likelihood function optimization, we also initialize the kernel of the HF model using the best hyperparameters from the LF model. This model seamlessly integrates derivative information, facilitating accurate reconstruction of general physical models by synthesizing data from multiple fidelities, including derivatives. To address computational demands, we incorporate kernel approximation techniques based on structured kernel interpolation with derivatives (D-SKI). The effectiveness of our proposed method is validated through three case studies: (1) solving a partial differential equation (PDE) describing fluid or scalar transport in a 1D domain, (2) reconstructing a high-resolution local turbulence velocity field using the Johns Hopkins forced isotropic turbulence dataset, and (3) predicting the surface of a rugged mountainous area while finding the peak via Bayesian optimization (BO). Comparative analysis against four other algorithms underscores the superiority of our approach across multiple fields and highlights its significant potential for engineering applications.

A comprehensive multi-fidelity surrogate framework based on Gaussian process for datasets with heterogeneous responses

Residual Gaussian Process: A Tractable Nonparametric Bayesian Emulator for Multi-Fidelity Simulations

Multi-fidelity Gaussian process surrogate modeling for regression problems in physics

Multi-Fidelity Surrogate Based on Single Linear Regression

A novel multi-fidelity surrogate modeling framework integrated with sequential sampling criterion for non-hierarchical data

A generalized hierarchical co-Kriging model for multi-fidelity data fusion

A novel multi-fidelity surrogate modeling method for non-hierarchical data fusion

Multi-Fidelity Residual Neural Processes for Scalable Surrogate Modeling

A comprehensive framework for multi-fidelity surrogate modeling with noisy data: a gray-box perspective

A multi-fidelity surrogate modeling method in the presence of non-hierarchical low-fidelity data

Multi-fidelity Hierarchical Neural Processes

Multi-Response Heteroscedastic Gaussian Process Models and Their Inference

Fast multi-fidelity gaussian processes with derivatives for complex system modeling

Multi-fidelity surrogate with heterogeneous input spaces for modeling melt pools in laser-directed energy deposition

A survey on multi-fidelity surrogates for simulators with functional outputs: unified framework and benchmark

Multi-fidelity data fusion for the approximation of scalar functions with low intrinsic dimensionality through active subspaces

Multi-fidelity information fusion with hierarchical surrogate guided by feature mapping

A Bayesian neural network approach to Multi-fidelity surrogate modelling

A Single-Fidelity Surrogate Modeling Method Based on Nonlinearity Integrated Multi-Fidelity Surrogate

A Latent Variable Approach for Non-Hierarchical Multi-Fidelity Adaptive Sampling

Multi-fidelity reduced-order surrogate modelling