Quantitative analysis of factor sensitivity of soil frost heave rate based on optimized WOA-BP strategy
Soil frost heave is one of the important challenges faced by engineering construction in cold regions.Accurate prediction and sensitivity analysis of soil frost heave are very important to ensure engineering safety and stability and prevent structural deformation and damage.Aiming at three kinds of soil samples in a mining area in Northeast China,the frost heave experiment of one-way freezing without water replenishment was carried out under different conditions of cold end temperature,moisture content and dry density.Based on the experi-mental data,the key factors affecting frost heave rate are deeply analyzed,and a WOA-BP prediction model with dry density,moisture content,cold end temperature,specific gravity and icing temperature as input varia-bles is constructed.On this basis,Chebyshev chaotic mapping and adaptive weight adjustment strategy are in-troduced,and the WOA-BP neural network with adaptive weight of Chebyshev chaotic mapping is optimized.It is verified that the model has small prediction error and can better predict the frost heave rate of soil.Combined with Garson algorithm,perturbation method and Monte Carlo simulation,the sensitivity analysis of the influen-cing factors of soil frost heave rate is carried out.The results show that the results obtained by all methods are consistent,namely the sensitivity of frost heave rate of soil samples in this mining area to the changes of dry density,specific gravity,moisture content,cold end temperature and freezing temperature decreases in turn.