Experimental study and prediction model on frost heave and thawing settlement deformation of subgrade soil in alpine meadow area of Qinghai-Tibet Plateau

Frost heave deformation and thawing settlement deformation of highway subgrade are common phenomena in cold regions. Freeze–thaw deformation causes uneven settlement of the subgrade and driving hazards. Thus, it is necessary to evaluate and predict the freeze–thaw deformation of subgrade soil. Based on the results of indoor freeze–thaw experiments, we establish a back-propagation (BP) neural network to predict subgrade soil frost heave and thawing settlement deformation. Freeze–thaw tests show that the frost heave rate and thawing settlement coefficient of the subgrade soil are positively correlated with the fine particle content, initial water content, and the number of freeze–thaw cycles (FTCs). The thawing settlement coefficient of the subgrade soil decreases with an increase in the dry density of the soil. A comparison of the predicted and measured values indicates a relatively small residual sum of squares (RSS) of the frost heave rate (0.027) and thawing settlement coefficient (0.315). The prediction model has high precision and is suitable for predicting the frost heave rate and thawing settlement deformation of subgrade soil. This research improves our understanding of the freeze–thaw deformation characteristics and settlement of highway subgrade soil.