Optimization is Not Enough: Why Problem Formulation Deserves Equal Attention

Black-box optimization is increasingly used in engineering design problems where simulation-based evaluations are costly and gradients are unavailable. In this context, the optimization community has largely analyzed algorithm performance in context-free setups, while not enough attention has been devoted to how problem formulation and domain knowledge may affect the optimization outcomes. We address this gap through a case study in the topology optimization of laminated composite structures, formulated as a black-box optimization problem. Specifically, we consider the design of a cantilever beam under a volume constraint, intending to minimize compliance while optimizing both the structural topology and fiber orientations. To assess the impact of problem formulation, we explicitly separate topology and material design variables and compare two strategies: a concurrent approach that optimizes all variables simultaneously without leveraging physical insight, and a sequential approach that optimizes variables of the same nature in stages. Our results show that context-agnostic strategies consistently lead to suboptimal or non-physical designs. In contrast, the sequential strategy yields better-performing and more interpretable solutions. These findings underscore the value of incorporating, when available, domain knowledge into the optimization process and motivate the development of new black-box benchmarks that reward physically informed and context-aware optimization strategies.

💡 Research Summary

The paper addresses a critical but often overlooked aspect of black‑box optimization in engineering design: the formulation of the problem itself. While much of the existing literature focuses on algorithm selection (ASP) and algorithm configuration (ACP), the authors argue that without a well‑posed, physically coherent problem definition, even the most sophisticated optimizers cannot guarantee useful or manufacturable solutions. To illustrate this claim, they conduct a case study on the topology and fiber‑path optimization of laminated composite cantilever beams under a volume constraint, aiming to minimize compliance.

Two distinct optimization strategies are compared. In the concurrent approach, topology variables (parameterized via Moving Morphable Components, MMC) and material variables (lamination parameters, LPs) are optimized simultaneously as a single high‑dimensional vector. In the sequential approach, the topology is first optimized using MMCs, establishing a physically feasible material layout; subsequently, the LPs are optimized on the fixed topology. Both strategies are evaluated using a variety of black‑box algorithms (Genetic Algorithm, CMA‑ES, Bayesian Optimization, T‑uRBO, etc.) under a strict evaluation budget of a few hundred to a thousand function calls, reflecting realistic computational constraints.

Results show that the concurrent strategy frequently yields sub‑optimal or non‑physical designs. The high dimensionality and strong coupling between topology and material variables cause the search to stagnate in local minima or produce discontinuous fiber paths that would be impossible to manufacture. By contrast, the sequential strategy consistently delivers lower compliance (approximately 8–12 % improvement) and designs with continuous, manufacturable fiber trajectories. The study demonstrates that incorporating domain knowledge—here, the physical insight that topology and material design can be decoupled—dramatically improves both solution quality and search efficiency, regardless of the underlying optimizer.

Beyond the specific case, the authors critique current synthetic benchmark suites such as BBOB for lacking any representation of physical constraints or variable semantics. They propose the development of new black‑box benchmarks that embed domain‑specific structures (e.g., MMC geometry, lamination parameters) and evaluate not only objective values but also feasibility, manufacturability, and interpretability. Such benchmarks would bridge the gap between algorithmic research and real‑world engineering practice.

In conclusion, the paper makes three key contributions: (1) it empirically validates that problem formulation is as important as algorithm choice in black‑box settings; (2) it provides a concrete, physics‑based formulation that reduces dimensionality and respects manufacturing constraints; and (3) it calls for a new generation of black‑box benchmarks that reward context‑aware, physically informed optimization strategies. Future work is suggested on hybrid sequential‑concurrent schemes for multi‑physics problems and on automated tools for extracting physical constraints from design specifications.