Speaking of Simulation Part 3 – Optimisation

This exclusive article series explores how leading organisations in the engineering simulation community utilise various simulation techniques. For each article in the series, we have interviewed key members of the NAFEMS community. In this instalment, the focus is on engineering optimisation—an area that is evolving rapidly, influencing multiple industries, and reshaping the way products are designed and developed.

At its core, engineering optimisation involves the systematic and automated exploration of a design space to find the best possible solutions that meet defined criteria. This may involve improving performance attributes, minimising weight or cost, and simultaneously addressing multiple objectives and constraints.

The interviews show that optimisation methods are widely applied in sectors such as automotive, aerospace, defence, and energy. They are employed throughout various stages of product development, from conceptual design to detailed engineering, addressing attributes like structural performance, aerodynamic efficiency, cost, and weight. Several interviewees have highlighted the importance of both commercial and bespoke software solutions, emphasising that tool selection often depends on the uniqueness of the task at hand.

Choosing the most appropriate optimisation technique hinges on factors such as problem complexity, model fidelity, and the stage in the design process. For instance, broad structural changes might call for topology optimisation in the early conceptual phase, whereas later refinements may involve tuning thickness or shape parameters. Engineers must also considerpractical aspects, such as familiarity with certain tools, the associated costs, and the stability and reproducibility of algorithms.

Recent advancements are broadening the scope of optimisation and making it more accessible and efficient. Multifidelity approaches, integrating both low- and high-fidelity models, can reduce computational costs while maintaining acceptable accuracy in surrogate model training. New software capabilities like automatic differentiation are enabling real-time optimisations on previously infeasible high-fidelity models. The growing trend of "democratising" optimisation involves providing pre-defined workflows and intuitive interfaces, thus allowing a wider range of engineers to benefit from sophisticated simulation toolchains without extensive specialist knowledge. Additionally, improvements in the visualisation and better methods for articulating user preferences are poised to enhance decision-making and foster truly user-centric design processes.

This article appeared in the January 2025 issue of BENCHMARK.