Generating and predicting soil desiccation cracking patterns utilizing a stochastic model based on geometric parameters

Abstract

Soil desiccation cracks and crack networks significantly influence the mechanical properties of soils. Accurate modeling and prediction of crack development are essential for both laboratory research and practical applications in geotechnical engineering and environmental science. In this study, a Desiccation Crack Simulation Program (DCSP) was developed on the MATLAB platform to simulate the evolution of soil desiccation cracks. Based on comprehensive statistical analyses of crack network images from previous studies and detailed observations of crack propagation, we propose a stochastic crack network generation model informed by geometric parameters and crack development processes. The model encompasses five key steps: (1) selection of crack initiation points, (2) crack propagation and intersection, (3) termination of crack growth, (4) secondary crack generation, and (5) final network formation. Key parameters include crack step size, randomized propagation direction, number of initial development points, and crack attraction distance. The DCSP enables both the rapid generation of random crack networks and the prediction of partially developed networks. The program was validated using two soil types, Xiashu soil and Pukou soil, demonstrating its effectiveness in simulating crack evolution. Prediction accuracy improves as crack network develops, highlighting the model's potential for predicting soil desiccation crack patterns.