Koen van der Blom
Koen van der Blom is our Leiden Institute of Advanced Computer Science's nominee for the C.J. Kok Jury Award 2019.
‘By developing mixed-integer algorithms that are able to handle multiple objectives, it becomes possible to find balanced solutions for a greater class of problems.’
About Koen's research
Buildings consume many resources, both during construction and during their lifetime. By optimising the design of a building, both of these expenditures can be reduced. In this case, the aim is to optimise a building spatial design. That is, the design of the exterior shape of a building as well as the internal division into different rooms. Reducing the resource requirements for construction can be achieved by minimising the strain on the structure of the building. The energy consumption during the building’s lifetime can be reduced by minimising the energy used for heating and cooling.
Since there are an endless number of possible alternative spatial designs, evolutionary algorithms are used to efficiently search through the many options for some good solutions. However, there is another problem. When we minimise the strain on the structure, this leads to completely different solutions than when we minimise energy consumption. Unfortunately, neither of these solutions is well balanced. To find balanced solutions that perform well in both, multi-objective optimisation is used. By providing a building engineer with a few high-quality solutions that each strike a different balance between the two objectives, the engineer can choose which design best suits the needs of this particular building.
The algorithms presented in this thesis stand out by tackling multiple optimisation challenges at once. Firstly, multi-objective problems are considered. Secondly, constraints are handled in order to ensure feasible designs. Finally, the use of mixed variables makes it possible to search through both whole numbers, fractional numbers, and categorical values.