Abstract: Particle shape is one of the main factors affecting the compactness, mechanical and seepage properties of sandy soils.
Fourier description is one of the mathematical methods effectively characterizing the grain shape. Based on the Fourier coefficient and
grain mean radius, we define the Fourier descriptor Dn. Employ Fourier descriptor D2, D3, D8 and, the logarithmic linear combination of
D3 and D8, we develop a simplified algorithm to automatically generate complicated soil grain shapes. The grain shape is controlled by
both Fourier descriptor Dn and phase angle δn series. The correlation between Fourier descriptors D2, D3, D8 and the grain geometrical
parameters. The results show that the length-width ratio a is positively correlated to D2, circularity C is negatively correlated to D2 and
D8, roundness R is strongly negatively correlated to D8, sphericity S is strongly negatively correlated to both D2 and D3 and, regulation Re
is negatively correlated to D8 and, negatively correlated to D3 when D8 equals to 0. Using the typical grains in the binary images of
Nanjing silty sand, obtained by micro-CT scanning, as samples, the shape similarity is calculated based on Pearson correlation
criterion. The correlation coefficient is higher than 0.94, suggesting that the algorithm possess the relatively high characterization
accuracy. The algorithm can be used to automatically generate massive practical sand grains in the two-dimensional numerical
simulation.

Key words: sandy soil, particle shape, Fourier descriptor, phase angle, similarity analysis