Study on prediction model of strength deterioration of moraine soil under high frequency freeze-thaw cycles
To explore the influence of freeze-thaw cycles on the strength of moraine soil,this study established a prediction model of the strength deterioration of moraine soil under different numbers of freeze-thaw cycles.Tak-ing the moraine soil in the western high-cold and high-altitude area as the research object,a total of more than 180 triaxial samples with a diameter of 101 mm were made by combining the indoor high-frequency freeze-thaw cycle test with the unconsolidated undrained triaxial test.The strength test was carried out after the high-frequen-cy freeze-thaw cycle test with different frequencies.The test variables were the number of high-frequency freeze-thaw cycles,the initial water content of the sample,and the confining pressure of the triaxial test.The influence of the number of high-frequency freeze-thaw cycles,the initial water content,and the confining pressure on the failure strength of the moraine soil sample was mainly explored.A BP neural network prediction model with three layers of input and one layer of output was established by MATLAB software.The data obtained from the experiment were imported into the BP neural network model for learning and training.The strength degradation prediction model of moraine soil under high frequency freeze-thaw cycle was established.At the same time,the SSA algorithm was introduced to improve the model.The number of neurons,the number of neural network lay-ers and the number of iterations in the BP neural network were optimized to improve the prediction accuracy of the model.The experimental results show that freeze-thaw cycles have a great influence on the strength of mo-raine soil.Under different testing conditions,the strength of moraine soil degrades in the range from 30%to 40%,and tends to stabilize after 15 to 20 freeze-thaw cycles.Based on the SSA-BP neural network model,the complex nonlinear relationship between the failure strength of moraine soil and the number of freeze-thaw cy-cles,initial water content and confining pressure can be reflected.The prediction error is significantly smaller than that of the BP neural network model,and the prediction results are more accurate.Comparing the SSA-BP neural network model with the Logistic model used in linear fitting,it is found that the two are in good agree-ment,which further illustrates the accuracy of the SSA-BP model.Through later verification,it is found that the output error of the model is small and has certain application prospects.This research can provide important pa-rameters for engineering stability evaluation in western cold and high-altitude areas.
moraine soilhigh frequency freeze-thaw cyclefailure strengthSSA-BP
NETL
NSTL
万方数据
