Tectonic labor: AI will increase the safety of construction in earthquake-prone regions

Russian scientists have created a neural network that analyzes data from seismic sensors and builds a map of underground structures based on them. Thanks to the new algorithm, it does this more accurately and faster than existing analogues. The development can be used to assess the risks of construction in earthquake-prone regions. For example, in the Far East, where the most powerful earthquake in 70 years occurred at the end of July. According to experts, such an analysis is necessary for the construction of any buildings, and in areas of geological faults it should be carried out with extreme care. The use of such technologies helps to avoid disaster and save money for construction investors.

How AI makes underground maps

MIPT specialists have developed an algorithm based on artificial intelligence that analyzes deep exploration data and independently creates detailed maps of underground structures. This technology makes it possible to accurately assess seismic risks during construction in regions with complex geology, where the probability of earthquakes is high. The technique may be in demand in the Far East, the Caucasus, as well as in the northern regions for monitoring the state of permafrost.

— At the input, the algorithm receives a response from the studied seismic environment, which is recorded during field work using various sensors, and at the output we receive a description of this environment according to various parameters. This way we can find out the boundaries of geological sections, identify inhomogeneities and assess where the foundation of a building can be placed without fear of rock deformation," Andrei Stankevich, an assistant at the Department of Computer Science and Computational Mathematics at MIPT, a graduate student, told Izvestia.

Photo: IZVESTIA/Pavel Volkov

To "look" into the bowels of the Earth, scientists use principles similar to echolocation: they send elastic vibrations tens of kilometers deep. By the nature of the reflection of these vibrations from different rocks, a map of the location of structures is created. Diffusion models are widely used for this task (a type of mathematical models that are often used in the work of neural networks). However, in order to restore a clear picture of the subsurface, they require hundreds or thousands of requests to AI, the MIPT said.

Miscalculation: the inspection of microelectronics was assigned to a robot inspector with AI

How automatic complexes can help improve the quality of functions and the cost of final products

Scientists have proposed a new technique based on the diffusion equation. It reduces the required number of calculations by several times — up to 50-100 calls with comparable quality, which significantly speeds up the process. Increasing the speed of work is crucial for creating high-resolution underground maps that require complex calculations.

— It's as if you saw underground structures through a fogged-up window, but step by step you gradually erased this fog, revealing a clear picture of what is hidden underground. This is how our system works: at the entrance we have a blurred model and a raw recording of seismic waves. Using a neural network, the system refines the map step by step and adds details to it. The output is a detailed map that corresponds to a realistic geological structure," said Andrei Stankevich.

Photo: IZVESTIA/Anna Selina

The method has already been tested on large sets of computer-generated data. It reconstructs maps more accurately than previous diffusion models, and it does so 10-20 times faster, while preserving subtle geological features. The algorithm can also be used to explore minerals.

Consideration of risks in earthquake-prone regions

As Grigory Agranov, a junior researcher at the Geological Institute of the Russian Academy of Sciences, explained, the assessment of the seismic hazard of the territory before construction should be carried out in any region, and in seismically active zones with special care.

— Such surveys are necessary during the construction of any building. Engineering geology should be carried out even during the construction of cottages, as there are groundwater and sands that are being washed away. In seismically active zones, the location of faults is very important. Catastrophic consequences are also possible in permafrost zones. All this is very important. Especially in difficult karst and seismically active zones," he told Izvestia.

Neural networks are now being actively used for these purposes, so improved software will be in demand in practice for geological and engineering surveys, the specialist added.

Photo: IZVESTIA/Sergey Konkov

The development of MIPT is theoretically capable of ensuring budget predictability for construction investors, reducing project time and reducing the likelihood of accidents due to more accurate planning, said Natalia Kruglova, Managing Partner of Trophy Assets.

— Currently, in earthquake-prone regions, design is carried out according to the old scheme: drilling, measurements, seismic exploration, and weeks for calculations and model construction manually or in basic programs. The method is too slow and does not scale well to large infrastructure projects. The indicators stated by MIPT are in the achievable zone. For example, in the Hiroshima University study, the AI system generates a map of the site based on seismic parameters in 10-60 minutes, which is commensurate with the goals and scale of acceleration announced by Russian developers," said Natalia Kruglova.

Gaze Program: AI simulating human eye will find cracks in bridges

A drone with video cameras will replace the work of at least three specialists

However, it is very important to avoid mistakes in such calculations, as they can lead to serious consequences. Therefore, a monitoring and verification system is needed: constant data verification, adaptation of algorithms to different geological conditions, and the availability of backup decision-making procedures. It is important to regulate and consolidate such technologies at the state level, the expert is sure.

The Far East is considered one of the most seismically dangerous regions of Russia. On July 30, the region experienced the most powerful earthquake since 1952 with a magnitude of 8.8. As previously reported by "Izvestia", according to calculations by seismologists, not all the energy accumulated in the earth's crust was released during this event, therefore, new tremors are likely to occur within a few months. The latest earthquake of magnitude 5 occurred 158 km from Petropavlovsk-Kamchatsky on August 10.

Photo: Global Look Press/Muhammad Fauzy

The regional authorities are closely monitoring the seismic situation: the Sakhalin Region cooperates with all relevant structures, including the Russian Academy of Sciences, in order to predict earthquakes and tsunamis. Currently, the region has a system for notifying citizens about emergency situations, the regional government said.