Jun.Prof. Thomas Wortmann

New Junior Professor Thomas Wortmann: Building sustainably using AI

To make buildings more material- and energy-efficient and thus contribute to combating global warming, there is no getting around computer-based design methods. Junior Professor Thomas Wortmann, a new tenure track professor for “Computing in Architecture” at the Institute for Computational Design and Construction (ICD), is investigating what these methods can look like. In his research, he relies on Artificial Intelligence (AI) and machine learning.
[Photo: Norbert Barth]

To an outsider, the fact that a student of architecture focuses on programming seems rather unusual. But that’s exactly what Thomas Wortmann did. After studying architecture in Kassel, he completed his master’s degree in Design and Computation at the Massachusetts Institute of Technology (MIT) and specialized in architectural computing. Already in the late 1960s, scientists at MIT were thinking about how computers could help design buildings; the first CAD software goes back even further.

Today, architectural computing is not only concerned with the design of buildings, but also analyzes buildings and simulates their energy use, for example. “Digitalization will irreversibly change architecture,” Thomas Wortmann believes. “To shape this change, students need to master computer science methods.”

Wortmann’s research takes the interplay between architecture and computer science to a higher level. “Architects use many methods from computer science to try things out. But I’m asking how and if these methods really work. The answers are then fed back into computer science, for which our questions are often too pragmatic.”

Future of us pavilion at the anniversary exhibition in Singapore. The shell construction, in which Thomas Wortmann's computer science methods were used, consists of 10,000 panels, each with a different degree of opening. This creates interesting lighting situations and - important in Singapore - the building manages without air conditioning.

A new research field is initiated

In his dissertation at the Singapore University of Technology and Design (SUTD), which was established in 2009 in cooperation with MIT, Wortmann showed what such a bridge between disciplines can look like in practice. In his work, he simulated and optimized the energy use of buildings in order to find the most energy-efficient solutions. Until then, it was assumed that evolutionary algorithms are best suited for such optimizations. “I found out that exactly the opposite is true,” Wortmann says with a trace of pride. In fact, model-based algorithms using machine learning were found to be more suitable for optimizing energy demand and, similarly, for optimizing daylighting. “There have hardly been any tests like this before, and I helped initiate this field of research.”

There are many areas of application for this in architecture. For example, if you want to find out what shape a high-rise building should ideally have in order to minimize wind load, Computational Fluid Dynamics (CFD) has so far been the method of choice. In CFD, problems in fluid dynamics are solved approximately using numerical methods. The problem with this is that a single computation can take several hours. However, solving optimization problems requires thousands of simulations. “This is not feasible,” says Wortmann.

Model-based algorithms allow to look at very many high-rise alternatives and select the one with the optimal wind load

Jun.Prof. Thomas Wortmann

This is where machine learning comes in. Here, a neural network is trained in such a way that it can predict the wind loads of different construction types. If a neural network is integrated into the algorithm, a model of the abstract space is created in which a solution is being sought. “Such model-based algorithms aren’t very accurate, but they are fast,” Wortmann explains. “They allow you to look at a large number of high-rise building alternatives and select the one with the optimum wind load.”

The next goal: multi-criteria solution spaces

Until now, computer-based design methods have looked at individual performance criteria such as wind load, energy efficiency, daylight quality, and material consumption. In the real world of construction, however, these factors are often interdependent. Wortmann’s research objective is, therefore, to create multi-criteria solution spaces in which several individual aspects are simulated simultaneously. “This task is very complex,” Wortmann describes the challenge. “Defining the problem, finding a solution, and understanding the solution - everything then becomes more difficult.” 

Performance Explorer. This software not only optimizes - in this example the shape of the pavilion on the left - but also represents the entire solution space, enabling architects and engineers to explore it interactively. Thus, they are not locked into one, "optimal" solution, but receive immediate feedback on the - in this case - static performance of different solutions through a machine-learned model.

The professorship fits well into the IntCDC Cluster of Excellence

He plans to include the execution of construction work more in these solution spaces in the future - in close coordination with Prof. Achim Menges and other leading researchers at the ICD in the field of robotic fabrication. This is entirely in the spirit of the Stuttgart Cluster of Excellence “Integrative Computational Design and Construction for Architecture” (IntCDC), where the professorship is based and whose main objective is the consistent digitalization of the design and construction process from planning to completion.

Thinking about what will happen on the construction site as early as in the planning stage, is uncharted territory in architecture. Also, new strategies are required to manage the exchange of data between the draft, statics, fabrication, and final assembly on the construction site. This is a problem that arises in a similar way between the different areas of the Cluster of Excellence. For this reason, Prof. Wortmann and Prof. Melanie Herschel from the Institute of Parallel and Distributed Systems (IPVS) at the University of Stuttgart have initiated a roundtable on “data strategy”, which is concerned with how to connect the various participants in a design and construction process on the level of computer science. A research proposal in cooperation with Prof. Steffen Staab, also from IPVS and Head of Analytic Computing, is to follow.

At the forefront of digitalization

In practice, however, architectural firms are often rather conservative, and for some the drawing pen is still a popular utensil. It will probably be some time before computer simulations become standard in the industry. “Currently, my research is aimed at those who are at the forefront of digitalization, but the architectural community will evolve to that point,” Wortmann anticipates. Especially since the tools are immediately useful in making buildings more efficient. One example is the software tool “Opossum”, an optimization tool developed by Wortmann that is based on machine learning. It is freely available and used by architectural and engineering firms worldwide. By now, you can conveniently get the tool from an app store. “We’ve already had more than 5,000 downloads.”

Eight years in Asia

Before his transfer to Stuttgart, Wortmann lived and worked in Singapore for six years and in Suzhou, China, for two years. In Asia, he learned to think big. Suzhou is “just” a suburb of Shanghai, but has seven million inhabitants. Elsewhere in Asia, too, rapid urbanization requires space for further millions of people. Population density makes it necessary to build upward and to look further into the future,” Wortmann believes. “That’s why Asians are more pragmatic and open-minded about new technologies.” The scientist wants to be carried by this spirit in order to make cities in Germany more climate-friendly and lower in emissions, because: “A sustainable future is a future in densely populated areas.”


Tenure-Track Prof. Dr. Thomas Wortmann, Institute for Computational Design and Construction, E-Mail
Tel.  +49 711 685 819 33                      

To the top of the page