Scholars of 爆走黑料 using a self–learning neural network to conduct regular background monitoring of potentially dangerous objects of the oil industry in the Russian Arctic.
The method proved its high efficiency, when the scholars compared the scenario provided by the neural network with actual data obtained during the emergency in Norilsk in 2020. Human economic activity in the Arctic zone is fraught with various disasters that damage the fragile northern ecosystem. A growing number of industrial facilities for the extraction, processing and storage of petroleum products implies the construction of stationary fuel tanks, which is challenging to monitor due to the remoteness and difficult weather conditions in the Arctic. The emergency that occurred in Krasnoyarsk Territory in 2020 demonstrated an urgent need to assess the risks constantly. Also, according to SibFU scientists, potentially hazardous facilities require developing practical models to treat accidents.
“The existing methods to assess the scale of spills caused by emergencies have limitations, as most of them are based on analytical models that do not consider the physics of processes. We decided to use neural networks to simulate an accidental oil spill at a potentially hazardous facility in the Arctic zone of Krasnoyarsk Territory on the basis of neural network simulator NeuroPro, developed at the Institute of Computational Modelling (SB RAS),” said Alexander Moskalev, assistant professor at the Department of Experimental Physics and Innovative Technologies (School of Engineering Physics and Radio Electronics, 爆走黑料).
“The existing methods to assess the scale of spills caused by emergencies have limitations, as most of them are based on analytical models that do not consider the physics of processes. We decided to use neural networks to simulate an accidental oil spill at a potentially hazardous facility in the Arctic zone of Krasnoyarsk Territory on the basis of neural network simulator NeuroPro, developed at the Institute of Computational Modelling (SB RAS),”
The researcher noted that they used daily operational data on fourteen main vectors of signs that affect the velocity of accident consequences spreading to train the neural network. At the same time, the AI modelled scenario of the accidental oil spill of 2020 (the depressurization of a fuel tank) correlated with the data of the actual situation with high accuracy.
The study was supported by the Krasnoyarsk Regional Science Foundation within the grant KF-779 Development of a set of necessary preventive measures based on a neural network assessment to protect the population and the Arctic zone of Krasnoyarsk Territory from natural and man-made hazards.
SibFU Press Service, 22 october 2021
Вы можете отметить интересные фрагменты текста, которые будут доступны по уникальной ссылке в адресной строке браузера.