SIBUR’s PolyLab R&D ecosystem is increasingly turning to artificial intelligence to control and predict the properties of materials and finished products, according to Artur Aslanyan, Head of Process and Service Development.
At the front end of innovation, AI tools are reshaping how new ideas are born. Instead of manually scanning vast volumes of global patents, scientific papers, regulatory texts, and industry news, researchers now rely on large language models to structure knowledge, identify high-potential polymer niches, and shift development from reactive problem-solving to predictive design.
Behind the scenes, SIBUR is also tapping into massive industrial datasets generated across the entire polymer value chain—from pyrolysis and synthesis to extrusion, lab testing, and final product processing. These data streams include production parameters, formulations, and performance results from PolyLab centers, feeding machine-learning models that map how composition and processing conditions determine final material properties.
One of the clearest demonstrations is a polypropylene film formula generator. Given target requirements such as strength, transparency, and elasticity—and accounting for specific production line constraints—the system outputs an optimized multilayer structure, including precise component ratios, layer thicknesses, and predicted performance characteristics.
A similar model is now being trained to forecast how recycled materials affect product color and quality. By analyzing historical data on recycled content and its blend with virgin polymers, the system aims to provide early-stage predictions of visual and quality outcomes—an area still being refined for higher accuracy.
Beyond materials design, SIBUR is also applying AI to product security. New systems under development aim to combat counterfeiting using embedded markers and digital tracking tools accessible across the supply chain, from processors to end users.
" We view artificial intelligence as a practical tool that is already changing our approaches to development at SIBUR, enabling us not only to accelerate processes but also to accurately predict the properties of the final product. The key challenges in implementing AI remain trust in model results, handling confidential data, and the quality of source information.
"To overcome these challenges, we are focusing on employee training, integrating AI tools into work routines, developing internal RAG solutions, and automating data collection," noted Artur Aslanyan , Head of Process and Service Development at SIBUR PolyLab.
The push is part of a broader digital transformation across SIBUR’s research divisions. At its Innovations center, the company is already testing AI-driven approaches to accelerate catalyst and material design.
By analyzing experimental datasets, models are uncovering relationships between molecular structure and performance, predicting catalytic efficiency, and reducing reliance on time-intensive lab work—moving the company toward faster, more controlled scientific development.
