Abstract: Simulating the mechanical behavior of loess in micro-scale is an effective way to reveal the mechanical mechanism of loess. Micro-mechanical model of loess is essential. At present, the microstructure of loess is often obtained by optical microscope or scanning electron microscope image, which, however, will lead to particle suspension and overlapping when directly used in two-dimensional modeling. In this paper, a method of randomly generating two-dimensional microstructure model of loess is proposed based on the cumulative distribution curve and the actual particle morphology of loess. Firstly, the cumulative particle distribution curve is divided into several intervals according to the particle size range, and the particle size is randomly extracted from each interval to form a particle size database which satisfies the size distribution. In subsequence, a database of particle morphology is established based on the microscopic images of loess particles. The size and morphology of particles are randomly extracted from the two databases to generate a particle bank with random sizes and morphology. Last but not the least, Monte Carlo method is used to randomly place these particles at the top of the model, hence a two-dimensional microstructure model of loess is established. The model established in this paper accords with the microstructure characteristics of loess and provides an initial model for the simulation and analysis of microscopic mechanical behavior of loess.

Key words: loess, microstructure model, Monte Carlo method, numerical simulation, scanning electron microscope