aiqbench analysis
This paper introduces a reduced-space multi-fidelity Bayesian optimization framework that significantly lowers computational costs for high-dimensional industrial process simulations by intelligently mixing low-cost approximations with expensive rigorous models. Validated on bioprocess and green fuel synthesis scenarios, the method demonstrates that adaptive sample allocation can drastically reduce evaluations of high-fidelity simulators without sacrificing optimization performance. The approach offers a scalable, simulator-agnostic solution for addressing the curse of dimensionality in complex black-box functions.
Reduced-Space Multi-Fidelity Bayesian Optimization of Process Simulation Models
arXiv:2609.17440v1 Announce Type: new Abstract: Optimizing industrial process flowsheets is often computationally prohibitive due to the high cost of rigorous simulations and the curse of dimensionality inherent in complex design spaces. To address these challenges, we present a reduced-space multi-fidelity Bayesian optimization (RS-MFBO) framework designed for high-dimensional, expensive black-box functions. The approach integrates Global Sensitivity Analysis (GSA) for dimensionality reduction with a fidelity-augmented Gaussian process that captures correlations between low-cost approximations and expensive high-fidelity evaluations. A cost-aware acquisition strategy, augmented with cooldown and promotion mechanisms, adaptively guides the allocation of samples across fidelities. The framework is validated on two distinct industrial process simulators: a plasmid DNA bioprocess in SuperPro Designer and a green fuel synthesis plant in Aspen HYSYS. Results across diverse economic and physical objectives demonstrate that the proposed method substantially reduces the number of high-fidelity simulator evaluations while maintaining competitive optimization performance compared to single-fidelity baselines. These results highlight RS-MFBO as a scalable, simulator-agnostic approach for cost-constrained black-box optimization.
Abstract:Optimizing industrial process flowsheets is often computationally prohibitive due to the high cost of rigorous simulations and the curse of dimensionality inherent in complex design spaces. To address these challenges, we present a reduced-space multi-fidelity Bayesian optimization (RS-MFBO) framework designed for high-dimensional, expensive black-box functions. The approach integrates Global Sensitivity Analysis (GSA) for dimensionality reduction with a fidelity-augmented Gaussian process that captures correlations between low-cost approximations and expensive high-fidelity evaluations. A cost-aware acquisition strategy, augmented with cooldown and promotion mechanisms, adaptively guides the allocation of samples across fidelities. The framework is validated on two distinct industrial process simulators: a plasmid DNA bioprocess in SuperPro Designer and a green fuel synthesis plant in Aspen HYSYS. Results across diverse economic and physical objectives demonstrate that the proposed method substantially reduces the number of high-fidelity simulator evaluations while maintaining competitive optimization performance compared to single-fidelity baselines. These results highlight RS-MFBO as a scalable, simulator-agnostic approach for cost-constrained black-box optimization.
Accepted at the 20th Learning and Intelligent Optimization Conference (LION 20), 2026. Corrected author version. This version corrects a typo in the mathematical description of the multi-fidelity covariance kernel in Section 3.2
Machine Learning (cs.LG); Optimization and Control (math.OC)
arXiv:2609.17440 [cs.LG]
(or arXiv:2609.17440v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2609.17440 arXiv-issued DOI via DataCite (pending registration)
Full text under the publisher's open licence · 210 words
