Abstract:According to the freezing and thawing test of three types of soil samples in the freezing project of Nantong metro, it is found that the frost heaving force, the frost heaving rate and the thawing settlement coefficient increase with the decrease of freezing temperature. Under the same temperature condition, freezing and thawing characteristic of the clay is the most significant, silty clay is the medium, and silt is the weakest. It is also found that under the comprehensive influence of soil type, temperature, water content and dry density, the freezing-thawing change rules of the three types of soil samples are also different, showing obvious uncertainty. Two different weight parameters are used to modify the excitation and output functions of the wavelet neural network, and the scaling and translation parameters are optimized by gradient descent method. On this basis, the prediction model of freezing-thawing characteristics of fuzzy random wavelet network was established with soil type, temperature, water content and dry density as input and freezing-heaving rate and thawing sedimentation coefficient as output. Engineering examples show that the predicted values of frost heaving rate and thawing sedimentation coefficient are basically consistent with the measured values of specific projects, which can be used as an effective tool to predict the freeze-thaw characteristics of underground freezing method in Nantong metro and surrounding areas.