摘要
为探究高频冻融循环对冰碛土力学性质的影响,并建立高频冻融循环作用下冰碛土强度劣化的预测模型,以高寒高海拔地区冰碛土为研究对象,通过室内高频冻融循环试验及不固结不排水三轴试验,探究高频冻融循环次数和初始含水率对冰碛土破坏强度的影响规律,并基于试验数据和BP神经网络建立相应的预测模型,同时引入群智能优化算法对模型进行改进。研究结果表明:(1)高频冻融循环作用对冰碛土强度有着很大影响,在不同试验条件下,冰碛土强度劣化范围为30%~40%,且经过15~20次高频冻融循环后冰碛土强度劣化趋于稳定。(2)利用SSA-BP神经网络模型可反映高频冻融循环次数、初始含水率、围压与冰碛土破坏强度之间复杂的非线性关系,且预测误差明显小于BP神经网络模型,预测结果更加准确。(3)将SSA-BP神经网络模型与线性拟合采用的Logistic模型对比验证发现两者吻合度较高,进一步说明了SSA-BP模型的准确性。研究成果可为高寒高海拔地区的工程稳定性评价提供重要参数。
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.Taking 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-frequency freeze-thaw cycle test with different frequencies.The test variables were the number of high-frequency freezethaw 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 layers 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 moraine 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 cycles,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 agreement,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 parameters for engineering stability evaluation in western cold and high-altitude areas.
作者
范金珂
岳祖润
韩子豪
孙铁成
胡田飞
张松
FAN Jinke;YUE Zurun;HAN Zihao;SUN Tiecheng;HU Tianfei;ZHANG Song(State Key Laboratory of Mechanical Behavior and System Safety of Traffic Engineering Structures,Shijiazhuang Tiedao University,Shijiazhuang 050043,China;School of Civil Engineering,Shijiazhuang Tiedao University,Shijiazhuang 050043,China)
出处
《冰川冻土》
CSCD
2024年第2期602-611,共10页
Journal of Glaciology and Geocryology
基金
中国国家铁路集团有限公司科研委托项目(P2021G047)
国家自然科学基金项目(52172347)
中央引导地方科技发展资金项目(226Z5402G)
河北省自然科学基金项目(E2023210064)资助。
关键词
冰碛土
高频冻融循环
破坏强度
SSA-BP
moraine soil
high frequency freeze-thaw cycle
failure strength
SSA-BP
