OpenLB-UQ: An Uncertainty Quantification Framework for Incompressible Fluid Flow Simulations Journal: ACM Transaction on Mathematical Software
Mingliang Zhong, Adrian Kummerländer, Shota Ito, Mathias J. Krause, Martin Frank, Stephan Simonis · 2025
Read OpenLB-UQ: An Uncertainty Quantification Framework for Incompressible Fluid Flow Simulations Journal: ACM Transaction on Mathematical Software by Mingliang Zhong, Adrian Kummerländer, Shota Ito, Mathias J. Krause, Martin Frank, Stephan Simonis (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-13867.txt.
Introduction
Mingliang Zhong and co-authors add an uncertainty module to OpenLB, an open-source library for lattice Boltzmann flow simulation. Simulations of fluid flow rest on input values that are never exact, so honest results need error bounds. Their 2025 paper, written for a mathematical software journal, builds that check into the library itself.
The module uses non-intrusive stochastic collocation, with polynomial chaos and Monte Carlo sampling of the inputs. Tests cover a Taylor-Green vortex with up to four uncertain inputs and flow past a cylinder. Convergence matched theory, and the method scaled well on large machines. Flow modelers get a toolkit that reports not just the flow, but how much to trust it.
Catalog record
| CREATOR | Mingliang Zhong, Adrian Kummerländer, Shota Ito, Mathias J. Krause, Martin Frank, Stephan Simonis |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 70 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2508-13867
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
OpenLB-UQ: An Uncertainty Quantification Framework for Incompressible Fluid Flow Simulations Journal: ACM Transaction on Mathematical Software Mingliang Zhong, Adrian Kummerländer, Shota Ito, Mathias J. Krause, Martin Frank, Stephan Simonis Abstract Uncertainty quantification (UQ) is crucial in computational fluid dynamics to assess the reliability and robustness of simulations, given the uncertainties in input parameters. OpenLB is an open-source lattice Boltzmann method library designed for efficient and extensible simulations of complex fluid dynamics on high-performance computers. In this work, we leverage the efficiency of OpenLB for large-scale flow sampling with a dedicated and integrated UQ module. To this end, we focus on non-intrusive stochastic collocation methods based on generalized polynomial chaos and Monte Carlo sampling. The OpenLB-UQ framework is extensively validated in convergence tests with respect to statistical metrics and sample efficiency using selected benchmark cases, including two-dimensional Taylor–Green vortex flows with up to four-dimensional uncertainty and a flow past a cylinder. Our results confirm the expected convergence rates and show promising scalability, demonstrating robust statistical accuracy as well as computational efficiency. OpenLB-UQ enhances the capability of the OpenLB library, offering researchers a scalable framework for UQ in incompressible fluid flow simulations and beyond. † † highlights: OpenLB-UQ: first open-source lattice Boltzmann method framework with built-in non-intrusive uncertainty quantification. Stochastic collocation generalized polynomial chaos achieves spectral convergence on flow around a cylinder and Taylor–Green vortex benchmarks. Hybrid sample- and domain-level parallelism scales efficiently, reducing UQ wall-clock time by a factor >30 . 1 Introduction Uncertainty quantification (UQ) is crucial in computational fluid dynamics (CFD), as it enables the assessment of the reliability and robustness of simulation results. CFD simulations are inherently affected by uncertainties arising from imprecise boundary conditions, variable operating environments, and uncertain material properties—factors that are rarely known with complete accuracy [ 1 ] . UQ methods provide the statistical foundation to propagate these uncertainties through simulations, transforming CFD from a purely predictive tool into a robust instrument for decision-making. Accurate treatment of uncertainty is fundamental for defining design margins, optimizing control strategies, and conducting effective risk assessments. Despite its importance and the increasingly efficient code developments in the last decades, open-source CFD frameworks that offer integrated support for advanced UQ techniques remain scarce. This paper fills that gap by presenting OpenLB - UQ, the first fully integrated, open - source lattice Boltzmann library that offers native support for state - of - the - art non - intrusive UQ techniques. Among the various branches of CFD, incompressible fluid flow simulations are widely used in engineering applications, including automotive, aerospace, and biomedical engineering. Incompressible flows characterized by constant fluid density and divergence-free velocity fields introduce unique challenges for UQ due to their pronounced sensitivity to input uncertainties. This sensitivity can significantly influence critical quantities of interest, such as velocity and pressure fields [ 2 , 3 ] , as well as derived aerodynamic forces like drag and lift coefficients [ 4 ] . Common sources of uncertainty in incompressible flow simulations include variations in geometric tolerances [ 5 ] , inflow boundary conditions [ 6 ] , and material properties such as viscosity and density. Accurately quantifying and propagating these uncertainties is essential for designing robust systems and for interpreting simulation results with greater confidence. Traditional CFD approaches typically involve the numerical solution of the governing equations of fluid flow most notably, the Navier–Stokes equations (NSE) to predict fluid behavior. While these methods are widely adopted, they can be computationally intensive and difficult to scale efficiently on high-performance computing platforms, particularly when UQ is incorporated. The lattice Boltzmann method (LBM) offers a mesoscopic alternative to traditional CFD techniques by simulating fluid behavior through particle distribution functions rather than directly solving the NSE [ 7 , 8 , 9 , 10 ] . LBM has gained considerable popularity due to its algorithmic simplicity, inherent parallelism, and effectiveness in handling complex geometries and boundary conditions. OpenLB, a continuously developed open-source LBM framework, provides comprehensive support for parallel computing through MPI, OpenMP, and CUDA, enabling efficient execution on a wide range of hardware from desktop workstations to supercomputers [ 11 , 12 ] . Its modular architecture and extensive model library have enabled diverse applications, including multiphase, turbulent [ 13 , 14 ] , thermal [ 15 ] , porous media [ 16 ] , and particle-laden flows [ 17 , 18 ] . These capabilities make OpenLB a robust platform for developing and validating novel methodologies—such as the UQ extension framework presented in this work. Monte Carlo sampling (MCS), often considered the benchmark for UQ, requires a large number of simulation evaluations due to its relatively slow convergence rate of \mathcal{O}(N^{-0.5}) , making its direct application computationally expensive for complex CFD problems [ 19 ] . To partially mitigate this issue, quasi Monte Carlo (QMC) method addresses this limitation by using low-discrepancy, deterministic sampling points [ 20 ] . Nevertheless, the convergence rate of QMC methods remains limited to approximately \mathcal{O}(N^{-1}) , which is still not sufficiently fast for many practical engineering scenarios. Consequently, alternative UQ approaches capable of significantly accelerating convergence are necessary. Generalized polynomial chaos (gPC) methods have emerged as attractive candidates because of their spectral convergence rates for smooth problems, dramatically reducing the required number of simulation evaluations. The gPC approach can be categorized into intrusive methods, such as stochastic Galerkin (SG-gPC), and non-intrusive methods, such as stochastic collocation (SC-gPC) [ 21 , 22 , 23 ] . While SG-gPC achieves highly accurate solutions, it requires extensive modifications throughout the underlying CFD solver, making its implementation challenging or infeasible within established simulation frameworks [ 24 ] . In contrast, SC-gPC is a non-intrusive, sampling-based method that requires minimal changes to existing solvers. Building on this advantage, we present a dedicated UQ extension for OpenLB that integrates SC-gPC along with other non-intrusive techniques. This extension enables efficient and scalable UQ within a high-performance lattice Boltzmann framework. By reusing a carefully optimized solver for each sample and minimizing additional overhead, our approach maintains solver modularity while supporting extensive UQ analysis. To evaluate the capabilities of the proposed OpenLB-UQ extension, we apply it to a series of representative test cases that demonstrate both fundamental correctness and practical applicability. The 2D Taylor–Green vortex, a canonical benchmark for incompressible flow solvers, serves to validate the implementation against its analytical solution. A variant of this case, incorporating four-dimensional uncertainty, further tests the framework’s ability to handle high-dimensional stochastic initial conditions. Additionally, the flow past a cylinder scenario evaluates the extension’s effectiveness in capturing uncertainty propagation in aerodynamic quantities such as drag coefficients and wake structures under stochastic inflow conditions. Collectively, these cases illustrate the robustness, accuracy, and versatility of the OpenLB-UQ extension in addressing a variety of UQ challenges. In addition, the framework can readily be used as a UQ-wrapper for any other application case in OpenLB, which includes a variety of complex fluid flow problems as well as multiphysics applications, both stationary and time-dependent in up to three spatial dimensions. Results across these test cases consistently validate the computational effectiveness and reliability of the SC-gPC based UQ framework integrated into OpenLB. By combining high computational efficiency with ease of integration, OpenLB-UQ provides a practical, fully open-source platform, eliminating the need for external UQ libraries