关键词: 土钉支护, 花岗岩残积土, 钉-土界面, 干湿循环, 剪切强度, 弱化累积, 神经网络

Abstract: Soil nailing reinforcement is a widely used reinforcement method for the granite residual-soil slopes in the Guangdong-Hong Kong-Macao Greater Bay Area. The region experiences frequent rainfall, and the interface between soil nails and granite residual soil is subject to wet-dry cyclic effects, gradually weakening the interface shear strength. This weakening effect accumulates with increasing cycles, leading to reduced slope stability. However, the cumulative weakening of the interface shear strength under long-term wet-dry cycles remains unclear. To address this, a series of small and large-scale direct shear tests was conducted under high wet-dry cycling conditions (up to 100 cycles) to analyze variations in cohesion, friction angle, shear strength, and other parameters of the nail-soil interface, considering the number of wet-dry cycles and interface saturation under consistent vertical stress conditions. The cumulative weakening effect of the nail-soil interface was quantified. Experimental results indicate that the apparent cohesion of the interface weakens rapidly during the first 10 cycles, but the rate of weakening stabilizes afterward. However, the total accumulated weakening remains significant. The friction angle of the interface did not exhibit a regular pattern of change as the number of wet-dry cycles increased. A neural network model was developed to predict the cumulative weakening of the shear strength of the nail-soil interface, and its accuracy was quantitatively evaluated and validated. The analysis reveals that the overall error of the neural network model is less than 10%, with low variability in prediction accuracy. The findings of this study provide theoretical support for assessing the long-term stability and predicting slope failure in soil nail reinforcement of the granite residual soil slopes in the Guangdong-Hong Kong-Macao Greater Bay Area.

Key words: soil nailing, granite residual soil, nail-soil interface, wet-dry cycles, shear strength, cumulative weakening, neural network.