This is a 4-year Fondecyt project that will make it possible to more accurately quantify the impact of uncertain scenarios on processes applicable to climate change, medicine, telecommunications, astronomy, development of smart materials, and mining.
Andrea Riquelme, Journalist.- A regular Fondecyt research fund was awarded to a UAI academic for the study of electromagnetic models in uncertainty scenarios with greater precision and speed. Carlos Jerez, PhD in applied mathematics and academic at the Faculty of Engineering and Sciences of the Adolfo Ibáñez Universitywill investigate for 4 years the development of mathematical and computational models for various applications of electromagnetism, a fundamental area of Physics and Engineering that seeks to understand and take advantage of the interaction between particles with electric and magnetic fields and that we use daily in millions of applications such as cell phones, Wi-fi, microwaves, radio or space exploration.
The study will deepen the development of advanced computational techniques leading to establishing design parameters and safety of operation for applications of electromagnetic signals as diverse as the use of radiation to eliminate a cancerous tumor or improve the electromagnetic compatibility of thousands of devices and even the efficiency of solar panels. Thus, being able to rigorously quantify the effects of uncertainty seeks to give greater realism to the simulations and allow adjusting these averages and ranges for decision-making in various matters of individual, collective and environmental impact, among others. This is of particular relevance since the “Preliminary Draft Standard for Electromagnetic Radiation Emission of Telecommunications Services” is being discussed today in the portfolio of the Ministry of the Environment.
Electromagnetic waves are used not only in telecommunications, but also in medicine, radar, or in estimating the amount of ore in a mining deposit. The project has components of mathematical modeling, high performance computing, data science and neural networks, tools with significant advances in the last decade, which will improve precision and reduce uncertainty.
Carlos Jerez, lead researcher of the project, comments that «In mathematical and computational models, deterministic values are assumed that in reality come from the average of experimental observations. Sources and parameters that have uncertain and complex behavior. This study will adjust the understanding of this randomness, which could be applied in a large number of studies that operate under these paradigms and thus obtain more precise and reliable results, as well as develop high-performance and cost-effective algorithms and computational codes.”
The specialist explains that until now researchers are systematically faced with the “curse of dimensionality” which forces them to generate a series of random numbers for each parameter and simulate over and over again, until obtaining a satisfactory average, which implies a high cost in time and hours of design without being able to really understand the effects of the probability distribution of those scenarios. This exercise is carried out in matters as diverse as climate forecasting, radiology, telecommunications, astronomy, development of smart materials and mining. «Quantifying the effects of uncertainty in the multiple applications of the electromagnetic field could be a revolutionary change with a high impact in R+D+i, providing a platform capable of dealing with decision management in highly complex, realistic and uncertain scenarios» says Jerez.
Uncertainty quantification is a recent branch of engineering that combines applied mathematics and computer science to provide a complement to those models that work with experimental or uncertain data.
The teacher José Pinto national co-investigator, doctor in electrical engineering and academic of the Faculty of Engineering and Sciences UAI, points out: “From an algorithmic point of view, the methods for problems in relatively few dimensions (5 or less) are impractical for problems of 20 or more, due to memory and/or time limitations. On the other hand, the standard methods for high dimensions, although they allow simulating various problems, their performance is still insufficient in many situations, for example, it is not possible to perform simulations in real time. Given this framework, the only alternative is to develop algorithms specialized to the problems under consideration that generate a significant improvement. For this, although knowledge of the aforementioned problems is necessary, a share of innovation is also required that allows the development of new algorithms or the import of ideas little explored in this context.
For Ryan McClarren, a professor in the College of Engineering at the University of Notre Dame, this project is key “for governments, companies and non-profit organizations to make decisions based on data, they must understand the uncertainty in any analysis. This project will develop the tools to not only predict the future, but also to make sure that stakeholders take uncertainty into account.”
