NASA Utilizes Deep Learning to Monitor Solar Flares

2019-11-12

Solar flares may be a threat to human safety often seen in science fiction movies or novels, but there is scientific basis for this. Solar flares and other space threats not only have a serious impact on spacecraft and satellites, but also cause damage to Earth systems such as wireless communication and GPS navigation. However, accurate monitoring of solar flares requires very expensive costs, and the predicted results are often unreliable. Now, the Frontier Development Lab (FDL) of the National Aeronautics and Space Administration (NASA) has developed a new method based on deep learning, which allows researchers to more accurately capture the sun's extreme ultraviolet radiation.


In the past, this extreme ultraviolet radiation (or EUV) was the main cause of solar flares, usually monitored by recording the magnetic field or plasma distribution on the sun and inputting it into a physics-based model to predict EUV emissions. However, FDL used a different approach: researchers created an alternative measure of EUV irradiance by deep learning on solar images. The images of the sun are provided by the Solar Dynamics Observatory (SDO), a spacecraft launched by NASA in 2010 to observe the sun from a geosynchronous orbit around the Earth.The researcher of this model, Alexander Szenicer, said, "Our research shows how to train a deep neural network to mimic the instrument on the Solar Dynamics Observatory (SDO). By inferring the level of ultraviolet radiation that the sensor will detect based on what other instruments on the SDO are observing at any given time, we have demonstrated the ability to improve the safety rate of NASA space missions and enhance our ability to monitor space threats."


The results of the deep learning approach surpassed the current physics-based models, and as part of the evaluation results, the team developed new benchmarks for comparing predictions, hoping that these benchmarks will be useful in future research.



This research is the result of an eight-week summer research accelerator conducted by FDL, which is a collaboration between NASA Ames Research Center, SETI Institute, and many other private and public partners, including Google Cloud, Intel AI, IBM, Nvidia, HPE, etc. Last summer, the challenge was to develop an AI model to use SDO images to predict spectral irradiance.


This research has been published in Volume 5, Issue 10 of the journal "Science Advances", titled "Deep Learning Virtual Instrument for Monitoring Extreme Ultraviolet Solar Spectral Irradiance".


Paper link: https://advances.sciencemag.org/content/5/10/eaaw6548

RECENT AI NEWS

RECENT AI TOOLS