Joe Alxandersen receives prestigious grant to fold together time and space
Joe Alexandersen, associate professor at the Faculty of Engineering, is among this year's 39 recipients of the Independent Research Fund's Sapere Aude grants for young talented research leaders. With a grant of just over 6 million, he aims to address one of the biggest challenges in so-called computational morphogenesis. This can, among other things, make wind turbines more efficient and silent.
Are computers better designers than us humans?
According to Joe Alexandersen, associate professor at the Faculty of Engineering , the answer is yes. At least in many cases, he explains, if you combine computer simulations with mathematical optimization.
As a researcher, he is working on optimizing structures, including using so-called computational morphogenesis. And, as he says, that's just a fancy term for letting computers instead of engineers design components for everything from wind turbine blades to computer chips.
To this, there are many advantages.
"Through computer simulations, we can create better and more efficient components for all sorts of purposes. And we can create components that can be produced more intelligently with fewer materials and, therefore, with less energy," says Joe Alexandersen.
"Human intuition can be powerful, but it is also very limited by what it already knows. With the computer, you can start from scratch, and that can result in new solutions that we never would have imagined. Sometimes we can also learn something entirely new about fundamental physics."
However, the researcher emphasizes that the computer cannot do it alone.
"It is still humans who tell the computer what to do through algorithms. So, it doesn't become entirely free from humans."
From 100 to 10 days
Joe Alexandersen has just received one of the prestigious Sapere Aude grants of just over 6 million from the Independent Research Fund. With it, he and his team will address one of the biggest problems in computational morphogenesis: the waiting time.
"The project aims to reduce the calculation time for these design methods. I want to advance the field so that it can handle not only static technical problems but also time-dependent problems," he says.
Currently, computational morphogenesis is mainly used to create components that perform tasks unaffected by their surroundings or external factors over time. These components remain constant over time.
But in the real world, few things stay the same like that. The problem, however, is that if you want the computer to design something that changes over time, it takes a really long time to calculate.
"Normally, such a computer-driven optimization process might take 10 hours for static problems because you have to perform between 100 and 1,000 repeated simulations. But if the component changes over time, suddenly each simulation takes 10 hours, and the whole process can end up taking 100-200 days," says Joe Alexandersen.
"Until now, people - including myself - have simply accepted that this is how it is. You can start it, and then you have to wait for the time it takes. But in an industrial context, that's not feasible. It makes no sense to wait 100 days for a result. So, the goal of the project is to try to go from, let's say, 100 days to 10 days."
Space-time and supercomputers
The reason it takes so long for the computer to calculate is that the usual optimization methods are sequential.
The computer first simulates the structure and its behavior. Then, it figures out how to improve it and creates a new simulation. And it continues like this in a loop with many simulations until you have the finished optimized solution. For time-dependent problems, each simulation is much more extensive.
"You can compare the static case to taking a single picture. But if what you need to optimize changes over time, one picture is not enough. You need an entire movie, a long sequence of pictures. And since it took a long time just to get that one picture, well, it takes a really long time to get a movie," explains Joe Alexandersen.
To solve the problem, Joe Alexandersen and his colleagues will work with what is called spacetime descriptions. In simple terms, they will fold time and space together to create a single entity called spacetime.
"You could say that instead of having a movie consisting of a long series of pictures in a sequence, we squeeze the movie into a lump that can be cut into many small pieces. Each of these pieces can then be sent to individual computers as part of a supercomputer, and each computer records, for example, 10 seconds of the movie at the same time, but each on its own. So, the whole movie takes only 10 seconds to record. It is also called parallel computing," says Joe Alexandersen.
Making wind turbine blades better
In the project, Joe Alexandersen will specifically work with a particular air nozzle as a case study.
The nozzle can be used on wind turbine blades to make the blade cut through the air more efficiently and quietly. And since it causes the air to move in irregular waves behind it, it is a good example of a component that changes over time solely due to its physics.
"It's an old design that hasn't really changed for a long time. My idea is that we can use computational morphogenesis to create a nozzle that is even better, and that might require it to look completely different," says the researcher.
But the project's hope goes far beyond just improving that one nozzle.
Joe Alexandersen hopes that by solving the problem of the long waiting time, they can make topological optimization and computati0nal morphogenesis useful for many more applications. This will ultimately mean that all kinds of products, machines, and devices that surround us will become better.