The calculation of seismotectonic deformations for different depth levels 1–15, 16–35, 36–70 km was performed according to the data of 1819 mechanisms of earthquake foci that occurred in Central Asia (φ = 25° – 60° N, λ=60° – 115° E) for the period from 1976 to the end of July 2020 with M>4.7. The orientation of the main axes of the strain tensor reconstructed from the mechanisms of earthquake foci with M>4.7 coincide at different depth levels with mainly submeridional and north-eastern shortening and varying elongation from sublatitude to north-western and near-vertical. The consistency of the orientation of the main axes of shortening and elongation reconstructed from seismological materials and from the published results of calculating GPS observations, is traced.

Manual processing of large volumes of continuous observations produced by local seismic networks takes a lot of time, therefore, to solve this problem, automatic algorithms for detecting seismic events are used. Deterministic methods for solving the problem of detection, which do an excellent job of detecting intensive earthquakes, face critical problems when detecting weak seismic events (earthquakes). They are based on principles based on the calculation of energy, which causes multiple errors in detection: weak seismic events may not be detected, and high-amplitude noise may be mistakenly detected as an event. In our work, we propose a detection method capable of surpassing deterministic methods in detecting events on seismograms, successfully detecting a similar or more events with fewer false detections.

This article discusses a new approach to processing lateral scanning logging while drilling data based on a combination of three-dimensional numerical modeling and convolutional neural networks. We prepared dataset for training neural networks. Dataset contains realistic synthetic resistivity images and geoelectric layer boundary layouts, obtained based on true values of their spatial orientation parameters. Using convolutional neural networks two algorithms have been developed and programmatically implemented: suppression of random noise and detection of layer boundaries on the resistivity images. The developed algorithms allow fast and accurate processing of large amounts of data, while, due to the absence of full-connection layers in the neural networks’ architectures, it is possible to process resistivity images of arbitrary length.

An original method has been developed for estimating formation dip and strike from transient induction LWD data, based on focusing in the time domain. The focusing consists in decomposing the measured signals into a time series and diagonalizing the matrix of focused magnetic field components. We have implemented the method and comprehensively tested it in horizontally-layered media used for LWD data inversion to solve geosteering problems and evaluate the formation resistivity. Estimates of the angles contribute to reliable geosteering when choosing a direction of drilling, as well as when inverting data for a complex earth model. A significant reduction in the resource intensity of inversion and model equivalence is achieved by reducing the number of determined parameters.