QUEENS: A Python Framework for Solver-Independent Multi-Query Analyses of Large-Scale Computational Models
Please login to view abstract download link
The systematic analysis of large-scale, high-fidelity computational models has become a central challenge in both research and industrial engineering, particularly in applications such as parameter identification, uncertainty quantification, Bayesian inverse problems, and scientific machine learning. These analyses typically require repeated model evaluations under varying parameter configurations, the integration of real-time data, or the generation of extensive simulation datasets. QUEENS (Quantification of Uncertain Effects in Engineering Systems) [1] is an open-source Python framework that enables such analyses by addressing both methodological and computational challenges. It provides a solver-independent infrastructure for composing, executing, and managing a wide range of multi-query analyses with arbitrary forward solvers, and it can be deployed on platforms ranging from local workstations to distributed high-performance computing systems. QUEENS serves as a research platform for developing and disseminating cutting-edge methodologies, with a particular emphasis on Bayesian probabilistic analysis. Besides, QUEENS offers a comprehensive portfolio of established methods, including design of experiments, gradient-based and gradient-free optimization, global sensitivity analysis, surrogate modelling, forward uncertainty quantification, and Bayesian inverse analysis. Complementing these algorithmic capabilities, QUEENS reduces the engineering overhead that typically dominates multi-query studies—job handling, parallel execution, data management and I/O, logging, and result aggregation—allowing researchers to focus on modelling and algorithms. QUEENS has already been applied across diverse computational disciplines, demonstrating its flexibility for both forward and inverse uncertainty-aware analyses. Recent studies have used QUEENS to estimate unknown model parameters from data, to propagate uncertainties efficiently through large finite-element models, and to identify which inputs most strongly influence a system’s response. Applications span nonlinear multiphysics problems in biomechanics, fluid–structure interaction, and porous‑media modelling. These examples showcase QUEENS as a robust, solver-agnostic engine for scalable probabilistic modelling, digital twinning, and data-informed scientific machine learning workflows. [1]: Biehler et al., arXiv:2508.16316 (2025).
