The safety of operating high-tech industrial facilities and engineering structures is one of the paramount tasks of modern science and technology. Currently, to ensure this safety, in addition to traditional regular diagnostics of the facility’s condition, systems of experimental monitoring and so-called “digital twins” of facilities are being created.
Employees of the Institute of Continuous Media Mechanics of the Ural Branch of the Russian Academy of Sciences, part of the Perm Federal Research Center of the UB RAS, have applied a new approach for identifying areas of increased load on a body’s surface based on deformation data, utilizing neural networks.
Doctor of Physical and Mathematical Sciences, Professor of the Russian Academy of Sciences Rodion Alexandrovich Stepanov: “As a rule, damage or destruction of an object arises from the prolonged impact of extreme environmental conditions, or as a result of rapid processes such as impact or explosion. For the latter case, none of the existing methods allow not only for real-time assessment of the structure’s condition but even for sufficiently prompt analysis.”
“In our work, we investigated the problem of determining the magnitude and, most importantly, the location of an unknown impact based on the structure’s responses recorded at only a few specific points. This is the so-called inverse problem, which is the most mathematically challenging and requires enormous computational resources when solved by classical methods. The new approach, involving artificial intelligence, allows us to overcome the most critical limitations.”
To implement this approach, a dataset for deep machine learning was created, on the basis of which an artificial neural network was developed capable of solving the problem of finding external loading in fractions of a second. According to the study’s authors, no other method can achieve such a result. Furthermore, the calculation time can be further reduced if the model is directly transferred to a microcontroller and becomes part of the sensor system.
About the Author
