| The freeze-thaw disease of subgrade in seasonal frozen soil area is becoming more and more serious,which affects the safe operation of road.The results show that the frost heave of subgrade is influenced by initial water content,temperature,compactness and freeze-thaw time.In order to determine the main factors affecting the frost heave characteristics of subgrade in seasonally frozen soil area,the prediction of subgrade deformation is carried out on this basis.The paper takes Xindeuqiao area in Kangding as the research area,and studies freeze-heave characteristics of subgrade in seasonal frozen soil area by using indoor subgrade model tests under the conditions of different water content,compaction degree,freezing temperature and freeze-thaw cycles.The prediction models of GA-SVR,PSO-SVR and SSA-SVR for the frost heaving characteristics of subgrade in seasonal frozen soil area are established to find out the most adverse factors that cause the frost damage of subgrade.The results can provide experimental and theoretical basis for the prevention and control of subgrade freezing disaster in seasonal frozen soil area.The main work and conclusions of this study are as follows:(1)Indoor subgrade model tests were conducted to analyze the freeze-heaving characteristics of subgrade under varying conditions,including water content,compacting degree,freezing temperature,and freeze-thaw cycles.Moisture,temperature,and displacement at the top of subgrade were considered.The results indicate that subgrade is more susceptible to freeze-thaw damage when the moisture content is high and the number of freeze-thaw cycles is greater.As the moisture content and number of cycles increase,the subgrade experiences greater deformation and faster deformation speed,and its freeze-resistance performance worsens.(2)Range analyses were found that the most adverse combination of subgrade frost heaving deformation is W3C2T1N3,that is,22%moisture content,65%compaction,-20℃freezing temperature and the fourth freeze-thaw cycle,the subgrade will produce the maximum frost heaving deformation.Variance analyses were performed to assess the influence of water content,freezing temperature,freeze-thaw cycles,and compacting degree on the freeze-heave of subgrade soil.The results show that water content has a very significant influence,freezing temperature and freeze-thaw cycles have a significant influence,and compacting degree has no significant influence.The order of the factors by influence was:water content>number of freeze-thaw cycles>freezing temperature>compaction degree.(3)Three prediction models(GA-SVR,PSO-SVR,and SSA-SVR)were established and evaluated using mean square error and goodness of fit.The PSO-SVR model demonstrated the smallest error,highest accuracy,and best fitting effect and was selected as the optimal model.The support vector machine model based on multi-algorithm optimization can feasibly predict the frost swelling characteristics of subgrade in seasonal frozen soil areas.(4)The sensitivity of each factor to the prediction model was analyzed using the PSO-SVR model.The results show that data disturbances in the input variables of the four factors were found to impact the output results of the prediction model.Specifically,the moisture content had the greatest influence on the maximum frost heave of the subgrade,followed by the number of freeze-thaw cycles,freezing temperature,and compacting degree. |