Translation. Region: Russian Federation –
Source: Peter the Great St. Petersburg Polytechnic University –
An important disclaimer is at the bottom of this article.
The 12th Artificial Intelligence Seminar was held at Peter the Great St. Petersburg Polytechnic University. The heads of the KNTN-3 project, "Artificial Intelligence for Solving Cross-Industry Problems," presented interim research results. Students, faculty, and researchers interested in this topic also participated in the seminar.
KNTN-3 is one of three key scientific and technological areas dedicated to the creation of digital platform solutions for multimodal data analysis (in accordance with the SPbPU Development Strategy to 2030). The area is led by Yuri Fomin, Chief Designer and Vice-Rector for Research at SPbPU.
At the seminar, Irina Anikina, Associate Professor at the St. Petersburg Polytechnic University School of Nuclear and Thermal Energy and Head of the "Flexible Power Plant Equipment Lifecycle Management System Using Predictive Analytics Tools" project, presented a prototype of the system. The project involves developing self-parameterizing digital twins of thermal power plants (TPPs)—dynamically updated models that take into account the actual condition of the equipment, its degradation, and technological limitations. This enables analyzing trends in key parameters reflecting equipment degradation and transitioning from planned to predictive maintenance. The combination of physical models (digital twins) and neural network approaches (a multilayer autoencoder based on LSTM and Transformer) overcomes the shortcomings of each method individually and improves the accuracy of anomaly prediction.
Digital twins have already been developed for six combined heat and power plants in the Northwest region, and neural network technologies have been tested on a PGU-450T combined cycle gas turbine unit, processing 536 parameters in real time. This has enabled a reliable assessment of the current condition of the equipment and reduced the risk of unscheduled repairs.
The developers' plans for 2026–2027 include registering software as a result of intellectual activity (RIA), implementing the system at TGK-1 facilities, developing functionality in the system for optimizing repair schedules for power equipment, and expanding the event library for automatic equipment defect detection.
The project's preliminary results already demonstrate the effectiveness of integrating machine learning and digital twins to generate recommendations for managing energy infrastructure assets, directly impacting their reliability and performance.
Daniil Miroshnichenko, a specialist at the Gazpromneft-Polytech Scientific and Educational Center, presented the interim results of the project "Automation of Seismic Data Processing Using ANN" to seminar participants. The project manager is Ivan Zhdanov, Chief Engineer of the Laboratory for Digital Modeling of Underground Oil and Gas Reservoirs and Well-Test Analysis. The researchers developed algorithms based on Transformer-type architectures and convolutional neural networks. These solutions automate routine operations (such as seismogram interpolation and noise filtering), which traditionally require significant time and highly qualified specialists. This will help reduce the processing time and free up geophysicists' resources for more complex interpretation tasks.
Marina Bolsunovskaya, head of the Industrial Stream Data Processing Systems laboratory at the SPbPU Advanced Engineering School "Digital Engineering" and project manager for the "Digital Platform for Transport Systems Data Analysis Using Hybrid Artificial Intelligence" project, presented the universal digital platform "POLANIS" and a hybrid AI optimizer for transport systems at the seminar.
The universal POLANIS platform and ecosystem enables the integration of computing modules, calculation version management, input data configuration, and results analysis through customizable dashboards. The platform serves as the foundation for creating digital models in transportation, industry, and other fields. The platform and optimizer enable the transition to predictive management of transportation systems, coordinated infrastructure development, and the implementation of highly automated modes of transport.
The seminar confirmed that our strategy is working. We are moving from theory to real systems. Digital twins of thermal power plants, neural networks for geologists, AI optimizers for transport—these are no longer prototypes, but tools that are changing approaches in energy, mining, and logistics. The strength lies in hybrid solutions, where physical models are enhanced by artificial intelligence. The results speak for themselves: higher accuracy, lower risks, and new opportunities. We're not just researching—we're creating a technological standard for industry," commented Yuri Fomin, Vice Rector for Research at SPbPU, on the seminar's results.
Please note: This information is raw content obtained directly from the source. It represents an accurate account of the source's assertions and does not necessarily reflect the position of MIL-OSI or its clients.