Having the "Highly Cited Paper" label next to your scientific article means great prestige in the scientific community. Colleagues of the Institute of Automation and Computer Science of the Faculty of Mechanical Engineering, Jakub Kůdela and Radomil Matoušek, are now also enjoying this mark, which belongs to one per cent of the most cited papers in the world over the last decade. In addition, the former recently scored success in the competition at the prestigious World Congress on Computational Intelligence conference.
Among the most cited scientific papers in the world is the article Recent advances and applications of surrogate models for finite element method computations: a review, which was published in late 2022 in the journal Soft Computing. "This is a so-called research article, where my colleague Matoušek and I looked at the replacement models in optimization over the last ten years: what works, what doesn't, what are the open software that can be used, where are the blind spots. It is arguably the longest article I have written, it contains no picture and it is the most cited one. I guess I'll draw some conclusions from it," laughs Jakub Kůdela, one of the two authors of the Institute of Automation and Computer Science.
The article deals with the mentioned equivalent models in optimization. "The thing is that you have a simulator and you try to perform optimization on it, but this optimisation is complex or computationally intensive, which means it is very expensive. You can't call the simulator half a million times to find out the effective setup. You have to do it differently and smartly. Replacement models are used for this," explains Kůdela.
Although the topic sounds quite complicated and theoretical, the truth is that models are also widely used in industrial practice. "Typically, in the design of complex structures, such as ships, cars... For example, when you calculate the aerodynamics of a new car and you want to change the parameters and find the optimal solution, computing on simulators is really expensive, the calculations are super demanding. That is why approaches with substitute models are used," explains Kůdela.
The cited article ranked alongside sixty of BUT that have achieved the prestigious Highly Cited Paper label since 2014. "It is a nice feeling and a great joy, there probably won't be many articles like this in a lifetime. It is nice to know that our work is appreciated by someone and that people read it," adds Kůdela, who recently won third place in the competition at two international conferences, the IEEE World Congress on Computational Intelligence and The Genetic and Evolutionary Computation Conference. "These are the top conferences focused on evolutionary computing and computational intelligence in general. The task was to find the optimal system settings on the specified simulator, which was the allocation of the photovoltaic panel to the grid. The magic is in finding an algorithm that does not take extremely much time and delivers nice results. I signed up with an algorithm that worked based on differential evolution. It is an old algorithm, but very suitable for this task if used smartly," says Kůdela.
So, what is the secret of success? "In optimization, it is always the case that when you see a problem, you have a lot of "tools", or algorithms, that you can use for it. It is about finding how to use the structures of a given problem to make the best possible selection and algorithm settings. It takes experience to know where to reach. That is actually what I enjoy most about my job, optimization as such. And the breadth of disciplines that I can work with here at the faculty. Optimization can be used almost anywhere: I work with colleagues of the Energy Institute to optimize energy systems, together with process engineers we cooperate on optimization of waste collection or distribution services, and so on. FME is a great place for my professional life," concludes Kůdela.