The object of this study is the process of deorbiting the KazEOSat-1 spacecraft, which has completed its active service life in low Earth orbit. The main problem is the lack of an effective technique to deorbit KazEOSat-1, taking into account its technical characteristics, orbital parameters, and the need to minimize risks to the environment and other objects in orbit. As part of the work, a software model was built that takes into account the initial orbital parameters of the device, which are essential for planning and performing deorbiting maneuvers. The model is designed to accurately calculate the descent trajectory, taking into account the laws of celestial mechanics and the influence of atmospheric conditions. The optimal deorbiting strategy was selected based on an analysis of various methods for calculating orbital maneuvers aimed at reducing fuel consumption and minimizing environmental risks. This included a comparative analysis of existing approaches and the selection of the most suitable ones under the given mission parameters. The results of the simulation using precise modeling methods in the MATLAB software environment allowed us to determine the main deorbiting parameters, such as the altitude at which the maneuvers begin, the required velocity impulses, the total fuel consumption, and the expected time before entering the dense layers of the atmosphere. Based on the obtained data, practical recommendations were formulated for the KazEOSat-1 deorbit. The first stage, the active controlled deorbit, is carried out by operating the low-thrust engine and braking by the Earth’s atmosphere, allowing the spacecraft to descend from 758 km to 444 km in 2.5 days. The second stage, the passive uncontrolled deorbit, continues the descent to 103 km in 969 days, using only atmospheric braking. The third stage, the uncontrolled drop, begins after reaching 103 km and ends with a drop to the Earth in 834 seconds

The object of this study is the accuracy of announcer identification based on short utterances. To solve the task of speaker identification based on ultra-short speech utterances, a phoneme-by-phoneme approach to constructing voice models has been proposed within the framework of the study. The validity of this approach is based on the fact that short utterances usually contain a limited number of phonemes. In this regard, a hypothesis was put forward assuming that in order to increase the accuracy of announcer identification based on short utterances, it is necessary to analyze the sound of specific phonemes by different announcers. The experiments involved speech recordings of monosyllabic words with corresponding phonemes, on the basis of which, using the ECAPA-TDNN neural network architecture, announcer voice models were constructed. The experimental studies showed that voice models constructed based on the sounds of only one model provide higher announcer identification accuracy compared to generalized models constructed based on all speech sounds. It was also found that different phonemes provide different announcer identification accuracy. For example, with a speech signal duration of 2–3 seconds, the accuracy of announcer identification by the generalized model was 75 %. And the accuracy of announcer identification using a model built on the basis of only one phoneme "E", with the same input data, was 85 %, which is 10 percentage points higher than that of the generalized model

Wildfires are significant ecological and environmental disasters, impacting forests, ecosystems, climate stability and human life. The visible-spectrum imagery-based traditional wildfire monitoring system can fail to perform well in the presence of smoke, haze and low lighting. A number of machine learning and deep learning techniques have been proposed, but most of the studies do not provide an integrated Short-Wave Infrared (SWIR) optical-electronic payload framework along with an intelligent optimization technique. The objective of this research is to design an intelligent SWIR-based optical-electronic payload architecture for accurate detection and remote sensing of wildfire and Earth applications via deep learning and optimization techniques. The proposed framework is based on Sentinel-2 SWIR satellite data layers with wildfire and non-wildfire samples. To enhance the quality of the images and the representation of their spectral domain, the following preprocessing operations are carried out: resizing, image normalization, SWIR band extraction, and data augmentation. The following spectral feature extraction techniques are then used: burn area analysis, vegetation stress analysis, and thermal anomaly detection. The framework also incorporates SWIR optical payload design, electronic subsystem development and SWIR InGaAs sensor modeling. Finally, a Hybrid Convolutional Neural Network (CNN)–Residual Network 50 (ResNet50) model optimized by Grey Wolf Optimization (GWO) is used for wildfire classification and hyperparameter tuning. The proposed framework achieved an accuracy of 91.03%, precision of 91.27%, recall of 91.03%, and F1-score of 91.01%. The wildfire detection capability, classification robustness, and convergence performance were enhanced through the integration of SWIR spectral analysis, hybrid deep learning and GWO. The proposed framework offers an effective and trustworthy solution for intelligent wildfire monitoring and Earth remote sensing applications with enhanced spectral sensing and classification performance.