Study on the critical rollover conditions of trucks on curved highway segments under sand-accumulated road conditions based on LSTM.

Objective: Sand on highways changes friction and superelevation, increasing rollover and skid risks. This study explores how sand accumulation affects truck driving stability and predicts the critical threshold at which rollover may occur under different road conditions and load scenarios.

Methods: This paper is based on Trucksim simulations of truck driving conditions on the Umal Highway's sand-prone curved sections (with curve radii of 60/100/215/400 m) under sand accumulation. Combining simulation data and using Long Short-Term Memory (LSTM) neural network algorithms, it predicts the lateral load transfer ratio (LTR) of a six-axle truck on the test section.

Results: The LSTM algorithm outperformed others, with superior accuracy metrics (R² = 0.99644, MAE = 0.0050118, MAPE = 0.00026711, RMSE = 0.0063982). Sand accumulation is classified into thin and thick stages. The thin stage primarily affects road friction, while the thick stage increases curve superelevation. When the sand just covers the asphalt pavement pores and the thickness of the sand is more than 166 mm or more, the loading quality of more than 25 tons six-axle trucks are more prone to rollover, when the rollover speed and the normal road state rollover speed compared to significantly lower, compared with the standard speed limit, and the magnitude of the drop even up to 33%.

Conclusions: The impact of varying sand accumulation conditions on speed thresholds differs significantly. Failure to promptly adjust speed limits during sand accumulation events may lead to rollovers even when drivers adhere to standard limits. The findings provide critical guidance for sand-prone highway management, recommending adaptive variable speed limits based on real-time sand thickness and road conditions to mitigate desert-related safety risks.