From the Editor-in-Chief

I.V. Yarmoshenko

PREDICTIVE MODELS OF DATA CENTER ENGINEERING SYSTEMS BASED ON ARTIFICIAL INTELLIGENCE: A REVIEW OF ARCHITECTURES AND DATASETS

A. S. Butorova, V. I. Borisov, A. P. Sergeev, M. V. Ronkin, V. S. Bobakov, S. S. Ivanov, S. O. Polyakov, K. A. Ignatkov,
Y. Y. Chernyshov, А. V. Skorokhodov, A. I. Manzhurov

This review addresses the problem of failure prediction in data centers’ (DCs) critical engineering systems. Modern DCs play a pivotal role in digital infrastructure, enabling the continuous operation of cloud services, telecommunications, and enterprise systems. As key elements of digital infrastructure, modern DCs ensure the continuous operation of cloud services, telecommunications, and enterprise systems. The growing computational demand leads to increased energy consumption and greater strain on engineering subsystems – primarily power supply and Heating, Ventilation, and Air Conditioning (HVAC) systems. The inefficient operation of these systems not only raises operational costs but also increases the risk of DC failures. The present work provides a comprehensive review of contemporary approaches to building failure prediction models for DC critical engineering systems using artificial intelligence techniques. It examines predictive modeling approaches and datasets used for model training, collected from both real-world and simulated environments. Special attention is given to neural network architectures, including Recurrent Neural Networks (RNNs), Convolutional Neural Networks (CNNs), Long Short-Term Memory (LSTM), and generative adversarial networks (GANs). The review also highlights challenges such as the
scarcity of labeled datasets and the limited adaptability and interpretability of existing predictive models. The review’s findings can inform the development of intelligent monitoring and predictive maintenance systems for DCs.

STATISTICAL CHARACTERISTICS OF THE GAMMA FIELD IN THE TERRITORY OF THE EAST URAL STATE RESERVE: READINGS FROM PORTABLE MEASURING INSTRUMENTS

A. A. Ekidin, D. D. Desyatov, E. I. Nazarov, M. D. Pyshkina, K. L. Antonov, A. V. Pykhova, A. R. Zigangirov

This series of publications covers the results of field studies, from planning to assessing the current activity of radionuclides in the soils of the East Ural State Nature reserve. This article presents the planning and execution stages of the field studies, which yielded representative results characterizing the radioecological situation over an area of over 14,000 hectares. The article is based on field logs recording readings from calibrated gamma-field radiation measurement instruments. Only incorrect entries representing no more than 1.2% of the total data for 170 in situ measurement points and soil sampling points were excluded from the primary data. Each measurement point with fixed coordinates characterizes a single section of the reserve and its adjacent territory with an area of ~1.22 km2. A graphical visualization of the field survey planning results was performed to understand the representativeness of the obtained primary data. The measured parameters at each point include separate groups: ambient dose equivalent rate at the soil surface and at a height of ~1 m above the surface (nSv/h); and the specific activity of cesium-137 and natural radionuclides (Bq/kg). The statistical characteristics of each parameter group were determined: distribution shape, value ranges, mean, and median values. Correlations between the parameters and possible data samples localized at individual sites were considered.

SOME METHODS FOR ESTIMATING THE IMPORTANCE OF VARIABLES IN STATISTICS. II. MACHINE LEARNING METHODS

Е. А. Kabakova, V. G. Panov

Some machine learning (ML) methods used for variable importance estimation are considered. The paper provides an overview of several tree-based classification methods (CART, AdaBoost, Gradient Boosted Trees, Random Forest, Extra Trees, Rotation Forest) and compares peculiarities of their application to biomedical data analysis. Additionally, possible adaptations of the algorithms for processing datasets with imbalanced classes are proposed. 

EXPERIENCE OF THE ACTIVITY POLESSIE STATE RADIATION-ECOLOGICAL RESERVE

M. V. Kudin

The paper presents an analysis of the science-based approaches that form the basis of functioning of the territory of the Belarusian sector of the Chernobyl NPP alienation zone from the moment of its formation and development up to the present time. A number of
issues and proposals for the maintenance of the evacuation (exclusion) zone at the present
stage are defined. 

MODIFICATION AND DEVELOPMENT OF WATER, SOIL AND VEGETATION POLLUTION INDICES FOR ENVIRONMENTAL MONITORING OF COPPER MINE INFLUENCE ZONE

A. N. Medvedev

The article addresses some methodical questions connected with generalized pollution indices application in the mining enterprise environmental monitoring when assessing their environmental impact. Many large mines include a whole set of studies in their environmental monitoring programs to assess the impact on all natural components to provide full compliance with environmental norms as well as to show the audience their environmental responsibility and commitment to saving nature. In this regard, the article
assesses the prospects of generalized indices use when processing complex monitoring data for water, soil and vegetation obtained in the vicinity of copper mine. As a first step widely known weighted arithmetic water quality index was modified using enhanced set of parameters, contemporary Russian norms and corresponding hazard classes for drinking and natural water pollutants. Evaluation of the modified index using environmental monitoring data confirmed that it is an effective tool for comparing river water pollution levels at different sampling points in different years. Using the Safyanovsky copper mine (Middle Urals, Russia) as a case study, it was concluded that mine water discharge into the Rezh River has no noticeable impact on river water quality, which remained virtually constant over the three-year study period (2017–2019) that confirmed by summarized data from Roshydromet’s annual reports. On the next step a complex index of soil and vegetation pollution with copper and zinc was assessed. Based on the calculation results using soil and vegetation sampling data, it was found that this index can provide clearer identification of anomalous points for outlining the pollution zone from the waste rock dump, since its amplitude is 30 % greater than for any of individual indices for copper and zinc. In general, the use of environmental pollution indices represents a promising direction for development the methods for processing integrated environmental monitoring data at mining enterprises.They can be recommended to obtain additional generalized information for municipal authorities and local population, confirming the admissibility of an enterprise’s impact on the environment and its environmental responsibility.