Bayesian optimization as a flexible and efficient design framework for sustainable process systems

TL;DR

This paper surveys Bayesian optimization (BO) as a versatile design framework for sustainable process systems, where objective evaluations are expensive and noisy. BO combines a probabilistic surrogate (commonly a Gaussian process) with an acquisition function to efficiently navigate the design space and select informative samples, formalized as maximizing $f(\mathbf{x})$ over $\mathcal{X}$ using data $\mathcal{D}_k$. It highlights motivating applications in materials design, reaction design, process design, and control, and surveys emerging directions such as high-dimensional, multi-objective, constrained, and multi-fidelity BO, along with batch and safety-aware extensions. The authors identify key challenges and future research directions in applications, methods, and theory, including systematic benchmarking, simultaneous handling of multiple problem features, and developing suboptimality guarantees to guide practice.

Abstract

Bayesian optimization (BO) is a powerful technology for optimizing noisy expensive-to-evaluate black-box functions, with a broad range of real-world applications in science, engineering, economics, manufacturing, and beyond. In this paper, we provide an overview of recent developments, challenges, and opportunities in BO for design of next-generation process systems. After describing several motivating applications, we discuss how advanced BO methods have been developed to more efficiently tackle important problems in these applications. We conclude the paper with a summary of challenges and opportunities related to improving the quality of the probabilistic model, the choice of internal optimization procedure used to select the next sample point, and the exploitation of problem structure to improve sample efficiency.

Figures (1)

• Figure 1: Two key steps of Bayesian optimization for minimizing a function (i.e., maximising its negation). Left: regressing a surrogate model with uncertainty. Right: optimizing an acquisition function (subject to optional constraints).