A comprehensive comparison of neural operators for 3D industry-scale engineering designs

TL;DR

The paper tackles fair benchmarking of neural-operator surrogates for 3D industry-scale PDE problems. It standardizes six datasets spanning thermal, elastic, elasto-plastic, time-dependent, and CFD tasks and classifies architectures into branch-trunk, graph-based, grid-based, and point-based families. Architectural enhancements are proposed to enable fair comparison on parametric and free-form geometries, and a unified training/evaluation framework is used to assess accuracy, efficiency, memory, and deployment complexity. Results show that performance is problem-dependent, with DCON, S-NOT, FigConvNet, and Transolver excelling in different regimes; the work provides practical guidance for deploying neural-operators in industry and releases the datasets, code, and models.

Abstract

Neural operators have emerged as powerful tools for learning nonlinear mappings between function spaces, enabling real-time prediction of complex dynamics in diverse scientific and engineering applications. With their growing adoption in engineering design evaluation, a wide range of neural operator architectures have been proposed for various problem settings. However, model selection remains challenging due to the absence of fair and comprehensive comparisons. To address this, we propose and standardize six representative 3D industry-scale engineering design datasets spanning thermal analysis, linear elasticity, elasto-plasticity, time-dependent plastic problems, and computational fluid dynamics. All datasets include fully preprocessed inputs and outputs for model training, making them directly usable across diverse neural operator architectures. Using these datasets, we conduct a systematic comparison of four types of neural operator variants, including Branch-Trunk-based Neural Operators inspired by DeepONet, Graph-based Neural Operators inspired by Graph Neural Networks, Grid-based Neural Operators inspired by Fourier Neural Operators, and Point-based Neural Operators inspired by PointNet. We further introduce practical enhancements to adapt these models to different engineering settings, improving the fairness of the comparison. Our benchmarking study evaluates each model strengths and limitations in terms of predictive performance, computational efficiency, memory usage, and deployment complexity. The findings provide actionable insights to guide future neural operator development.

Figures (3)

• Figure 1: Branch-trunk model architecture enhancements for freeform geometry.
• Figure 2: Branch-trunk model architecture enhancements for freeform geometry.
• Figure 3: Geometric neural operator architectures incorporating parametric design inputs.