IBM and NASA to predict the future solar storms make a digital twin of the sun
Most sun Complex secrets can soon be resolved thanks to artificial intelligence. On August 20, IBM and NASA announced the launch of Surya, a basic pattern for The Sun. This artificial intelligence tool, using a large dataset of solar activity, aims to deepen humanity’s understanding of solar climate and precise predictions of solar flames – electromagnetic radiation emitted by our star, which threatens both astronauts in the infrastructure and communication infrastructure on Earth.
Syria, with nine years of data collected by the NASA solar Observatory (SDO), has been trained in a spinning tool since 2010 and used high -resolution images every 12 seconds. To estimate the temperature of the star layers, SDO records the sun’s observations at different wavelengths of different electromagnetic. There are also accurate measurements of the sun’s magnetic field – essential data to understand how energy moves through the star and predicts solar storms.
Historically, the interpretation of this wide amount of diverse and complex data has been a challenge for hliophysics. To address this challenge, IBM says Surya developers have used SDO data to create a digital twin from the sun – a dynamic virtual replica of the star that is updated when recording new data, and it can be manipulated and easily studied.
This process began with the unity of the various data molds fed in the model and allows it to process them continuously. Next, a long-range vision transformer was used-AI’s conventional analysis of very high resolution images and identifying the relationships between their components regardless of their distance.
The performance of the model was optimized using a mechanism called the gate that reduces the amount of memory consumption by up to 5 % by filtering the noise in the data, thereby enhancing the quality of the processed information.
More precise predictions in less time
Its developers say the project gives Surya a significant advantage: Unlike other algorithms that need to be labeled the extensive data that is fed, Surya can learn directly from raw data. This allows to quickly adapt to different tasks and provide reliable results in less time.
During the experiment, Syria showed its versatility to integrate data from other tools, such as the Parker solar probe and the solar observatory and the heliosphere (SOHO), two other spacecraft that see the sun. Syria also proved that it is effective in various predictor functions, including predicting flame activity and solar wind speed.
According to IBM, traditional prediction models can only predict one hour before an hour before the signals identified in specific areas of the sun. In contrast, “Syria, using visual information, provided two -hour guidance. It is thought to be the first to warn this type. In the initial experiment of the model, the team said that they had achieved 16 % improvement in the accuracy of solar flame classification, significant progress over existing methods.”
NASA emphasizes that although this model is designed for the study of heliophysics, its architecture is consistent with various contexts, from planetary science to Earth’s observation. “By preparing a fundamental model trained on NASA’s heliophysical data, we can analyze the complexities of the sun’s behavior with unprecedented speed and accuracy,” said NASA’s director of data science. “This model empowers a wider understanding of how solar activity affects the important systems and technologies that we all rely on the Earth here.”
The risk of abnormal solar activity is not partial. A large solar storm can directly affect global communications, collapse electrical networks, and GPS navigation disorders, satellite operations, internet connections and radio transmission.
Anders Moshoz-Jaramilo, a solar physicist at the South Western Institute of Research in San Antonio, Texas and the main scientist of the project, emphasized that Syria’s goal is to maximize lead time for these scenarios. “We want to give the longest possible time to the ground. Our hope is that this model has learned all the important steps behind our star transformation over time so that we can extract practical insights.”
This story appeared first Wired en español And translated from Spanish.
