Multi-objective optimisation of a Stator¶
Problem Description¶
The problem involved three objectives and 20 decision variables. The first formulation came from the engineer in the company. We applied our in-house BO algorithm to solve it. After several discussions, we found out it was not a 20 but nearly 50 dimensional problem because of an additional mechanical system in the optimisation process. But we did not step here. Another engineer suggested that he could simplify the problem and combine two systems (with physics magic) and use only 15 decision variables. We ended up with three objectives and 15 decision variables.
Why was tailoring needed?¶
Tailoring was needed in the problem formulation and not in the algorithm. The tailoring modified the problem formulation and we hoped that it would result in a better reflection of the real-world problem.
Baseline algorithm¶
In-house Multi-objective BO
Tailoring process¶
Mostly discussions after solving the first problem formulation. In the future, we may end up with a different problem formulation. Challenges: Different terminology, converting decision-makers' experience, knowledge and their needs into an appropriate optimisation formulation
What was tailored¶
No response
Main problem characteristics¶
- Multi-objective
- very expensive simulations
- medium dimensionality
- parallelisation
- hard black-box constraint handling
References¶
No response
Contact information (optional)¶
No response
Author¶
Tinkle Chugh No response