Модель параллельного вычислительного процесса для многогоризонтного прогнозирования неисправностей в условиях ограниченных ресурсов
Работая с сайтом, я даю свое согласие на использование файлов cookie. Это необходимо для нормального функционирования сайта, показа целевой рекламы и анализа трафика. Статистика использования сайта обрабатывается системой Яндекс.Метрика
Научный журнал Моделирование, оптимизация и информационные технологииThe scientific journal Modeling, Optimization and Information Technology
Online media
issn 2310-6018

A parallel computational process model for multi-horizon fault prediction in resource-limited conditions

Dahe A.,  Stuchilin V.V. 

UDC 004.94:004.8
DOI: 10.26102/2310-6018/2026.62.11.004

  • Abstract
  • List of references
  • About authors

The paper addresses multi-horizon fault forecasting for intelligent monitoring of engineering and industrial systems under computational and response-time constraints. The aim is to develop a coordinated computational method in which a common representation of the current system state is used to generate horizon-specific forecasting trajectories and support subsequent technical-state assessment. The proposed method combines shared input preparation with specialized forecasting trajectories: Seq2Seq-LSTM for 6- and 12-hour forecasting and Temporal Fusion Transformer (TFT) for 24-hour forecasting. After the common representation is formed, the trajectories are executed concurrently and their outputs are temporally aligned for unified system-state analysis. The method is experimentally compared with a sequential baseline using identical models, input data, and parameters on two hardware configurations: an NVIDIA Tesla T4 GPU and a CPU-only platform. The evaluation considers inference latency and forecasting quality. On the NVIDIA Tesla T4, concurrent execution reduces the critical model-inference time from approximately 22 ms to 17 ms, a reduction of approximately 23 %, while preserving the forecasting functionality of the considered trajectories. On the CPU-only platform, the sequential organization requires approximately 453 ms and the parallel organization requires approximately 585 ms, corresponding to a 29 % increase in the critical inference path. The results characterize the investigated computational configurations and provide a basis for low-latency multi-horizon monitoring under resource constraints, while highlighting the hardware-dependent nature of the observed latency reduction.

1. Dahe A., Stuchilin V.V. Usage of machine learning algorithms for fault detection and diagnostics in life support systems. Izvestiya Sankt-Peterburgskogo gosudarstvennogo tekhnologicheskogo instituta (tekhnicheskogo universiteta). 2024;(70):93–99. (In Russ.).

2. Mehdiyev Sh. On monitoring the technical condition and technological safety of functional elements of the cyber-physical infrastructure. In: Cybersecurity for Critical Infrastructure Protection via Reflection of Industrial Control Systems, 27–29 October 2021, Baku, Azerbaijan. IOS Press; 2022. P. 18–26. https://doi.org/10.3233/NICSP220029

3. Demidova L., Marchev D. Development of the forecasting model for the complex technical systems' failures time during the proactive maintenance using the recurrent neural networks' technology. In: Proceedings of the 2020 2nd International Conference on Control Systems, Mathematical Modeling, Automation and Energy Efficiency (SUMMA), 11–13 November 2020, Lipetsk, Russia. IEEE; 2020. P. 370–374. https://doi.org/10.1109/SUMMA50634.2020.9280781

4. Xiao X., Ma Z., Xu D., Liu J. Remaining useful life prediction method of product based on improved transformer and GRU. In: Proceedings of the 12th International Conference on Quality, Reliability, Risk, Maintenance, and Safety Engineering (QR2MSE 2022), 27–30 July 2022, Emeishan, China. IET; 2022. P. 1725–1729. https://doi.org/10.1049/icp.2022.3115

5. Gulomov Sh.R., Yusupov B.K., Kamilov Sh.Sh.U. Models and algorithms for solving problems associated with large amounts of data in the military sphere. In: Proceedings of the 2020 International Conference on Information Science and Communications Technologies (ICISCT), 04–06 November 2020, Tashkent, Uzbekistan. IEEE; 2020. https://doi.org/10.1109/ICISCT50599.2020.9351506

6. Benjelloun S., El Aissi M.E.M., Loukili Y., et al. Big data processing: batch-based processing and stream-based processing. In: Proceedings of the 2020 4th International Conference on Intelligent Computing in Data Sciences (ICDS), 21–23 October 2020, Fez, Morocco. IEEE; 2020. https://doi.org/10.1109/ICDS50568.2020.9268684

7. Pan H. Lightweight design and implementation of machine learning models in time series forecasting. International Scientific Technical and Economic Research. 2025;3(2):160–171. https://doi.org/10.71451/istaer2533

8. Zhadan A.A., Shynkarenko V.I. Modern methods and tools for working with time series. Science and Transport Progress. 2025;(4):142–154. (In Ukrainian).

9. Lim B., Arık S.Ö., Loeff N., Pfister T. Temporal Fusion Transformers for interpretable multi-horizon time series forecasting. International Journal of Forecasting. 2021;37(4):1748–1764. https://doi.org/10.1016/j.ijforecast.2021.03.012

10. García-Nava J.-L., Flores J.J., Tellez V.M., Calderon F. Fast training of a transformer for global multi-horizon time series forecasting on tensor processing units. The Journal of Supercomputing. 2023;79(8):8475–8498. https://doi.org/10.1007/s11227-022-05009-x

11. Lazcano A., Sandubete J.E., Jaramillo-Morán M.A. A comparative framework for multi-horizon time series forecasting: neural networks with adaptive preprocessing. Machine Learning with Applications. 2025;22:100781. https://doi.org/10.1016/j.mlwa.2025.100781

Dahe Ahmad

National University of Science and Technology "MISIS"

Moscow, Russian Federation

Stuchilin Vladimir Valeryevich
Candidate of Engineering Sciences, Docent

National University of Science and Technology "MISIS"

Moscow, Russian Federation

Keywords: computational model, multi-horizon forecasting, parallel computational process, computational optimization, fault prediction, limited resources, time series, intelligent monitoring, cyber-physical systems

For citation: Dahe A., Stuchilin V.V. A parallel computational process model for multi-horizon fault prediction in resource-limited conditions. Modeling, Optimization and Information Technology. 2026;14(11). URL: https://moitvivt.ru/ru/journal/article?id=2562 DOI: 10.26102/2310-6018/2026.62.11.004 .

© Dahe A., Stuchilin V.V. Статья опубликована на условиях лицензии Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NS 4.0)
81

Full text in PDF

Скачать JATS XML

Received 03.08.2026

Revised 27.10.2026

Accepted 10.11.2026