An algorithm was found for predicting extreme phenomena
The Greek researcher Themistocles Sapsis, an associate professor in the Department of Mechanical and Shipbuilding at MIT, together with his postdoctoral researcher Mohammad Farazmand, found an algorithm that helps predict sea levels.
Researchers have devised a way to identify in advance key signs that precede an extreme event.
The technique can be applied to a wide range of complex systems, which seem to emit their own warning messages, as long as one can “hear” them, according to the Athens News Agency.
It is no coincidence that the study of the Greek engineer is of interest to the US Armed Forces, so it is funded by the Research Offices of all three Arms (Navy, Air Force and Army).
“There is no method today to explain when these extreme events will happen. We applied our new method to chaotic fluid flows, which is the “holy grail” of extreme events. If we can predict these, then we hope we will be able to implement some control strategies to avoid these extreme events, ”said Sapsis.
Until now, attempts to predict extreme events have been based on solving dynamic equations with incredibly complex mathematical formulas to predict the evolution of a complex dynamic system over time. However, according to Dr. Clearly, the physics of several complex systems is still poorly understood and their modeling involves serious errors, so the corresponding mathematical equations are not realistic.
But, he says, even in systems where their physics are well understood, there is a huge number of initial conditions by which one can feed the relevant dynamical equations, resulting in an equally huge number of possible outcomes, which makes it almost the real prediction of an extreme event is impossible.
As the Greek engineer points out, “if we just blindly take the equations and start looking for initial conditions that will evolve into extreme situations, there is a very high chance that we will end up with initial conditions that are extremely exotic, in other words, they will never happen. in practice. So the equations contain more data than we really need. “
Alternatively, if you put aside the equations and try to look only at real-world data for characteristic patterns that would warn of an impending extreme event, he points out, it would take a huge amount of data over a very long time to be able to timely detect any such warning sign with some certainty.
Sapsis created a computer algorithm that combines both equations and real data. This combination helps better predict extreme events in the real world.
Testing the algorithm in a model simulating the chaotic fluid dynamics has shown that it is capable of predicting a future extreme event in 75% to 99% of the cases, depending on the complexity of the fluid. Chaotic fluids exist in many forms around us, from cigarette smoke and the flow of air around an airplane engine to the circulation of air in the air, ocean currents or the circulation of blood in the body.
Th. Sapsis graduated from the NTUA School of Civil Engineering in 2005 and received his PhD from the MIT Department of Mechanical Engineering in 2011, where he also did post-doctoral research. Since 2016 he is an Associate Professor in the same Department of MIT.