Computing power of Delft supercomputer also available for BK

Research is increasingly based on data and how data can be used to achieve better designs, decisions or measurements. This sometimes requires enormous computing power to process all models and scenarios. Soon all researchers, students and education at TU Delft will be able to use the computational power of DelftBlue to solve complex problems. This high-performance computer has a speed of no less than 2 petaflops (one million times one billion calculations per second). The Faculty of Architecture and the Built Environment will also have access to use this supercomputer from the end of April.

High-performance computing at BK – Delft AI Labs

The faculty is working on innovations that will be made further possible by the supercomputer.

Examples can be found in the three Artificial (AI) Labs that have started at BK, as part of the TUD-wide AI Labs programme. In these AI Labs, experts ‘In AI’ work together with experts from various application domains to advance AI within these applications. The three AI Labs in Architecture are:

3D Urban Understanding (3DUU)

Thanks to the developments in photogrammetry (gathering information from photos with an emphasis on dimensions) and 3D computer vision, we can now capture the built environment on a large scale in 3D. In the 3DUU Lab, we will develop new methods and techniques to automatically recognise and model objects in the built environment in 3D. The techniques will automatically add information about what each object or surface represents in the data, such as a building, a tree, a traffic light, or a terrain. Based on this information, we can obtain insight into the current state or the future state of a certain area of a city and apply it in various applications, such as planning, management, visualisation, analysis, and simulation.

AI for Design, Analysis, and OPTimization in Architecture & the Built Environment (AiDAPT)

At AiDAPT we develop AI methods to assist architects and engineers with analyzing large amounts of available data (e.g. designs, images, videos measurements of buildings and constructions) to further optimally integrate it into their complex decision-making process. From initial design to life-cycle planning: the creation and operation of the built environment entail a number of complex decisions and analyses. This complexity exceeds the capacity of existing computational methods to process available data and optimize decisions at large scales. AI provides us with unprecedented capabilities to make sense of, and guide our decisions in, immense spaces. In turn, the intricacies of architectural and structural engineering applications offer a fertile validation and development ground for machine learning methods towards domain-specific intelligence.

Activating Intelligence in Building Lasting and Liveable Environments (AiBLE Lab)

The built environment faces major societal challenges. Planners, architects, and citizens using or living in these spaces may all have different views and priorities – so any strategy for dealing with these challenges must incorporate different preferences into the design process.

The AiBLE Lab will investigate how AI can be developed and used in complex real-world scenarios. The aim is to help reach effective, transparent and lasting decisions and agreements. This means incorporating human feedback into the loop, iteratively improving decision-making and driving behaviour changes. We focus on two important built environment challenges: energy transition and circularity.

Education

Education at BK is also inseparable from high-performance computing such as the Computational Design Studio (BK7083), and the two elective courses offered by the MSc Geomatics, i.e., Machine Learning for the Built Environment (GEO5017), and Modelling wind and dispersion in urban environments (GEO5015).

DelftBlue

With the development of the supercomputer DelftBlue, the Delft Institute for Computational Science and Engineering (DCSE) meets the enormous need among researchers and students for high-end computing resources.

The goal of DCSE is to bring together all TUD researchers in the field of computational science and engineering and to intensify cross-faculty research. The education offered by DCSE is intended for both bachelor, master, doctoral students, postdocs and all staff members. The institute has been in existence since 2017. A second five-year term has been awarded at the end of 2021 and is supported by the six founding faculties and the Faculty of Architecture and the Built Environment which recently joined. Access to DelftBlue can be requested via the self-service portal.

More information

For our faculty, Professor Jantien Stoter is the contact person for both DCSE and DelftBlue. If you have any questions about this, please contact her.
3D Urban Understanding (3DUU), Leader: Dr. Liangliang Nan, Urbanism (and EEMCS)
AI for Design, Analysis, and OPTimization in Architecture & the Built Environment (AiDAPT), Leaders: Dr. Seyran Khademi (Architecture) and Dr Charalampos Andriotis (AE&T)
Activating Intelligence in Building Lasting and Liveable Environments (AiBLE Lab), Leader: Dr Tong Wang, MBE (and EEMCS)

If you want to know more about DCSE and DelftBlue, please visit the website of DHCP, where you can also register for the DCSE newsletter or watch the video TU Delft | The Delft High-Performance Computing Centre (DHPC)

Previously published news items: TU Delft supercomputer officially opened