2687-0517 18 2 2025 1-136

RAR RUS 7-20 Peter the Great St. Petersburg Polytechnic University Misharina Tatiana tanechkamisharina254@gmail.com

St. Petersburg, Russian Federation

0000-0003-0093-6506 Peter the Great St. Petersburg Polytechnic University Malov Sergey

St. Petersburg, Russian Federation

Categorical survival analysis of the required job execution times in the hybrid supercomputer center According to statistics, the actual execution time of most jobs on a supercomputer cluster differs significantly from the time requested by the user. Investigation of distributions of supercomputer job execution times using statistical or machine learning methods allows optimizing the operation of a supercomputer cluster. We study the results of computational jobs processing in the supercomputer center of Peter the Great St. Petersburg Polytechnic University. We have developed a nonparametric approach for detection and statistical confirmation of weak stochastic orders based on categorical nonparametric framework of contrasts obtained from the Kaplan–Meier estimators obtained from independent groups of right-censored observations. To adjust the confidence level of the detected weak stochastic orders, we apply the Bonferroni correction to all the comparisons under consideration. We perform comparative statistical analysis of the distributions of required execution times to complete successfully the job in different groups of right-censored observations; detect and confirm available weak stochastic orders. 10.18721/JCSTCS.18201 519.2 данные типа времени жизни оценка Каплана–Мейера критерий типа Вальда стохастические порядки суперкомпьютерный кластер планировщик задач https://infocom.spbstu.ru/article/2025.85.1/

RAR RUS 21-32 Saint-Petersburg Mining University Bazhin Vladimir bazhin-alfoil@mail.ru

St. Petersburg, Russian Federation

Engineering Laboratory LLC Anufriev Aleksandr anufriev_rf@yahoo.com

St. Petersburg, Russian Federation

Saint-Petersburg State Institute of Technology (SPSIT) Rusinov Leon

St. Petersburg, Russian Federation

System of interconnected solutions “Intelligent Quarry” The article discusses the implementation of digital solutions in the management system of the mining and transportation complex using the case of Karelsky Okatysh JSC. An analysis of the initial state of the enterprise’s technological chain is presented, highlighting key issues related to the stability of the blend composition, the quality of the mined ore and the efficiency of its transportation management. To address these problems, the software and hardware system called “Intelligent Quarry” was developed, comprising interconnected modules for blend stabilization, automated raw material quality monitoring and predictive equipment condition control. The effectiveness of the proposed solutions was confirmed by simulation of the expected results and subsequent comparison of production indicators before and after the system’s implementation. A technical and economic analysis confirmed the increase in concentrate output by 0.84%, a decrease in magnetic iron content in tailings by 0.2% and an additional annual concentrate yield of 171438 tons. 10.18721/JCSTCS.18202 622.001 mining and transportation complex intelligent mine mining and processing plant automated management systems hyperspectral sensing neural networks https://infocom.spbstu.ru/article/2025.85.2/

RAR RUS 33-44 Peter the Great St. Petersburg Polytechnic University Andrusenko Andrei andrusenkoau@gmail.com

St. Petersburg, Russian Federation

56049610600 0000-0003-1116-7765 Peter the Great St. Petersburg Polytechnic University Drobintsev Pavel drobintsev_pd@spbstu.ru

St. Petersburg, Russian Federation

Dataset creation for comprehensive performance evaluation of automatic speech recognition systems 10.18721/JCSTCS.18203 004.522,004.934 automatic speech recognition test dataset large language models punctuation and capitalization context-biasing https://infocom.spbstu.ru/article/2025.85.3/

RAR RUS 45-55 57225127284 0000-0001-9325-0356 ITMO University Tomilov Nikita programmer174@icloud.com

St. Petersburg, Russian Federation

58910796700 0009-0009-1470-7633 ITMO University Turov Vladimir firemoon@icloud.com

St. Petersburg, Russian Federation

A page-based approach for storing vector embeddings This study proposes a page-based approach to organize the storage for vector embeddings combined with the use of general-purpose lossless compression algorithms. The proposed approach organizes vector embeddings into pages of a configurable number of entries that contain vector embeddings and all necessary metainformation, and then the page files are compressed using general-purpose compression algorithms. This approach allows configuring page size and specific compression algorithm, to balance retrieval speed and storage efficiency. Experiments on three datasets, including PyEmb-50GB with more than 28 million dense vector embeddings, showed that the proposed solution reduces the occupied disk space by 14–40% compared to existing storage formats, such as ORC and Parquet, and up to two times compared to SQLite and H2. In addition, the suggested approach demonstrates a comparable to SQLite and H2 vector retrieval time, which is also a hundred times faster than ORC and Parquet. The results indicate that increasing the page size logarithmically reduces the storage size, while linearly increasing retrieval time. The proposed storage format supports thread-safe vector access, reducing both the necessary disk space and retrieval time, making it a robust solution for large-scale vector data management. It can also be used in approximate nearest neighbor search, provided the correct way of sharding vector embeddings between pages. 10.18721/JCSTCS.18204 004.42 vector embeddings compression of vector embeddings ORC Parquet https://infocom.spbstu.ru/article/2025.85.4/

RAR RUS 56-73 Skiba Vladimir bauman@bmstu.ru 0000-0003-0644-1731 Petrenko Sergei petrenko.sa@talantiuspeh.ru 0000-0002-8605-8865 Sirius University of Science and Technology Gnidko Konstantin

Krasnodar Krai, Russian Federation

0000-0002-9954-4643 St. Petersburg Electrotechnical University “LETI” Petrenko Alexei

St. Petersburg, Russian Federation

Concept of ensuring the resilience of operation of national digital platforms and blockchain ecosystems under the new quantum threat to security The obtained results in the field of quantum informatics clearly demonstrate the high technological potential of quantum technologies. A cryptanalytically relevant or significant quantum computer can threaten the operation of various systems, including national blockchain ecosystems and platforms in the Russian Federation. In this situation, a concept of ensuring the resilience of the operation of national digital platforms and blockchain ecosystems under the new quantum security threat is needed, the provisions of which are substantiated in this article. The concept contains a justification for the relevance of the problem and strategic goals of ensuring quantum resilience, national interests in the field of quantum information technologies, the presence of quantum threats to the operation of digital platforms and blockchain ecosystems, methods, means and priority measures to ensure the quantum resilience of national digital platforms and blockchain ecosystems. The main directions of further research of the group “Technologies for countering previously unknown quantum cyber threats” of the Scientific Center for Information Technology and Artificial Intelligence of the Sirius University of Science and Technology, aimed at implementing the proposed concept, are also considered. 10.18721/JCSTCS.18205 004.03 information security quantum information technology quantum security threat quantum resilience technological safety operational safety verifiable safety https://infocom.spbstu.ru/article/2025.85.5/

RAR RUS 74-90 Ivlev Vladislav nevidd@yandex.ru Nikiforov Igor igor.nikiforov@gmail.com 6603839750 Peter the Great St.Petersburg Polytechnic University Sergey M. Ustinov usm50@yandex.ru

Polytechnicheskaya, 29, St.Petersburg, 195251, Russia

IT project infrastructure setup automation with help of large language models This study conducts an analysis of existing large language models (LLMs) and AI agents, identifying Llama 2 as the most suitable model for automating IT project environment configuration. A mathematical model of the proposed method is introduced to automate IT infrastructure setup and reduce development time. The system architecture incorporates modules for natural language processing (NLP), configuration generation and command execution. The effectiveness of the method is evaluated through experiments across five key production scenarios, comparing two approaches: traditional infrastructure configuration tools and the proposed LLM-based method utilizing Llama 2. Experimental results demonstrate that the proposed method reduces configuration time up to 60%, decreases error rates from 25% to 8% and improves configuration quality approximately in 3 times. The article is relevant to IT professionals engaged in automating development and infrastructure configuration processes, as well as researchers exploring the application of artificial intelligence, particularly large language models, in the IT industry. 10.18721/JCSTCS.18206 004.89 Large Language Model Llama 2 AI agent IT infrastructure setup automation Natural Language Processing configuration generation artificial intelligence https://infocom.spbstu.ru/article/2025.85.6/ 74-90.pdf

RAR RUS 91-98 Pilipko M.M. m_m_pilipko@rambler.ru Morozov Dmitriy dvmorozov@inbox.ru Incremental delta-sigma modulator A delta-sigma modulator with reset for incremental ∆Σ ADCs for the 180 nm CMOS technology with a supply voltage of 3.3 V from Mikron JSC is presented. The simulation of the ∆Σ modulator in the time domain in the Virtuoso analog design environment from Cadence DS was performed. The clock frequency was set to 6.25 MHz. The power consumption was about 9.5 mW. The reset was performed every 32 or 128 clock cycles. The results of the ∆Σ modulator simulation were processed in MATLAB. The digital decimation filter in the form of a cascade of integrators was realized in software. At the oversampling ratio of 32, the modulator shows SINAD = 69.3 dB (ENOB = 11.2 bits) and SFDR = 76.9 dB. At the oversampling ratio of 128, SINAD = 88.7 dB (ENOB = 14.4 bits) and SFDR = 92.7 dB are achieved. The crystal dimensions were 640 x 340 µm. The ∆Σ modulator circuit is suitable for precise digitization of sensor signals in the audio frequency range. 10.18721/JCSTCS.18207 621.3.049.774.2 analog-to-digital converter delta-sigma modulator incremental delta-sigma ADC bootstrapped switch dynamic element matching https://infocom.spbstu.ru/article/2025.85.7/

SCO RUS 99-110 0009-0001-8226-6288 Mironov Kirill Morozov Dmitriy dvmorozov@inbox.ru Akhmetov Denis akhmetov_rphf@yahoo.com Stabilized reference current source for biomedical applications Neurostimulators are devices used to electrically stimulate the nervous system. They are a promising alternative to existing pharmacological methods of treating neurological disorders. The paper presents the current driver, one of the key blocks for providing electrical stimulation. The basic requirements and characteristics of this device are described. The noise of the reference current source, current mirror has been analyzed and the effect of the differential amplifier noise on the total noise current at the output devices has been considered. Based on the results of the analysis, a method for estimating the required output impedance of the current driver is proposed. The circuit implemented in 180 nm CMOS process. The output impedance of not less than 30 MOhm is obtained at the output current of 140 µA and the voltage compliance of 90.9% of the supply voltage. A comparative analysis of the results with the work of other authors is given. 10.18721/JCSTCS.18208 621.3.049.77 current source neurostimulation noise PSD output impedance feedback https://infocom.spbstu.ru/article/2025.85.8/

SCO RUS 111-119 Pilipko M.M. m_m_pilipko@rambler.ru Morozov Dmitriy dvmorozov@inbox.ru A pipeline analog-to-digital converter in 180 nm CMOS A pipelined analog-to-digital converter (ADC) is presented, which was designed using 180 nm complementary metal-oxide semiconductor (CMOS) technology with a supply voltage of 1.8 V from Micron JSC. The ADC circuit consists of a sample-and-hold device, an 8-level redundant stage, five 6-level redundant pipeline stages, a back-end 3-bit ADC, as well as synchronization circuits, an adder and multiplexers to get at the output the 16-bit direct binary code of the whole ADC or the redundant code from first to fifth stages. The pipeline is implemented as a switched-capacitor circuit using operational transconductance amplifiers. The simulation of the ADC in the time domain in the Virtuoso analog design environment from Cadence DS was performed. The clock frequency was set to 50 MHz. The power consumption was about 52 mW, the following main characteristics were achieved: SINAD = 74.6 dB (ENOB = 12 bits) and SFDR = 75.3 dB. 10.18721/JCSTCS.18209 621.3.049.774.2 analog-to-digital converter pipeline ADC bootstrapped switch time interleaving redundant stage https://infocom.spbstu.ru/article/2025.85.9/

RAR RUS 120-136 0000-0001-7126-6787 Zhemelev Georgiy 56049610600 0000-0003-1116-7765 Peter the Great St. Petersburg Polytechnic University Drobintsev Pavel drobintsev_pd@spbstu.ru

St. Petersburg, Russian Federation

A software system for surrogate-based prototyping of gas turbine blades using serverless containers in the cloud Design optimization of gas turbine blades is a complex multidisciplinary task requiring computationally expensive physics simulations. To perform them, a multitude of computer-aided engineering tools are used, often with machine-learning surrogates for rapid prototyping, all integrated into the optimization cycle. However, current approaches to such integration are hindered by the need for labor-intensive manual setups, vendor lock-in and a lack of scalable, automated workflows. We present a novel cloud-based architecture for building flexible optimization pipelines using containerized components. The proposed solution employs serverless containers, asynchronous messaging and cloud services to ensure the system’s scalability, portability and resilience. Additionally, it follows MLOps principles to achieve reproducibility and efficient lifecycle management of machine learning models used in the optimization process. Unlike existing frameworks, our solution minimizes user setup complexity, allows easy integration of various software into the optimization cycle, and avoids vendor lock-in through open-source technologies and standard cloud APIs. Experiments with aerodynamic design optimization of gas turbine blades demonstrate the system’s scalability, fault tolerance and successful integration of surrogate models for rapid blades prototyping. Furthermore, the system’s flexibility and extensible architecture make it applicable to a broader range of engineering design optimization tasks beyond gas turbine blade aerodynamics. 10.18721/JCSTCS.18210 004.896 gas turbine blades engineering design optimization serverless containers cloud computing surrogate models machine learning MLOps https://infocom.spbstu.ru/article/2025.85.10/