Smart solution for simulations
Research takes crucial step in efficient handling of time-consuming simulation models
How to find the optimal spot for each wind turbine in a wind farm without them getting in each other's way. Or how to accurately determine how large the holes in a factory filter should be to effectively block polluting particles. The key to solving these challenges often lies in sophisticated simulation models. However, there is a stumbling block: the time-consuming nature of these simulations.
Best fit
Laurens Bliek, researcher in the Information Systems group has taken an important step in dealing efficiently with time-consuming simulation models. His solution: the use of surrogate models. Bliek's research focused on understanding which surrogate models best fit specific problems and how this depends on the computing power required.
Mimicking simulations
Bliek: "By using AI to cleverly mimic simulation models, you can save a lot of computation time in complex optimization problems. Our software offers the opportunity to solve these problems, compare techniques and gain new insights about the surrogate models used in mimicking simulation models."
Publicly availableThe result of the research includes not only a software package, but also essential insights into problem properties, guidelines for the application of techniques and knowledge about mimicking simulation models with surrogate models. Making the software package and all the data collected publicly available contributes to the scientific community and the practical application of this innovative approach.
The research, conducted in collaboration with TU Delft's Algorithmics group, provides a valuable contribution to the research areas of Artificial Intelligence and Data Science at TU/e.
Read all about the research project: https://research.tue.nl/en/publications/benchmarking-surrogate-based-optimisation-algorithms-on-expensive