An intelligent model for predicting the behavior of soil conditions depending on external weather conditions

¹ Siberian Federal University, 79 Svobodny pr., Krasnoyarsk, 660041, Russian Federation
² Reshetnev Siberian State University of Science and Technology, 31, Krasnoyarsky Rabochy Ave, Krasnoyarsk, 660037, Russia

^* Corresponding author: oleslav@mail.ru

Abstract

The study focuses on the development of an intelligent system for monitoring and forecasting the condition of road surfaces in Eastern Siberia, addressing challenges posed by extreme climate fluctuations. Seasonal variations in temperature and soil moisture critically impact the load-bearing capacity of road structures, leading to accelerated wear, deformations, and safety risks. This research integrates advanced machine learning models, including LSTM, Transformer, TCN, and XGBoost, to predict changes in road conditions based on meteorological and soil data. Field measurements of soil elasticity modules were analyzed to assess seasonal impacts, with LSTM demonstrating the highest accuracy (MSE: 0.025, MAE: 0.0045). The findings confirm that freezing increases soil stability during winter, while spring thawing causes significant weakening due to over-saturation. Strengthening road bases with 30% sludge improved their durability and resilience under heavy loads. The proposed system combines real-time monitoring with predictive analytics, offering a practical tool for infrastructure management in extreme climates. Key outcomes include optimized maintenance schedules, recommendations for spring traffic restrictions, and strategies to mitigate road degradation. This work highlights the potential of machine learning in enhancing the efficiency and safety of road infrastructure, contributing to sustainable transportation in cold regions.