The complex behavior of granular material considering large deformation and post-failure is of great interest in the geotechnical field. Numerical prediction of these phenomena could provide useful insights for engineering design and practice. In this paper, we propose a novel numerical approach to study soil collapse involving large deformation. The approach combines a recently developed critical state-based sand model SIMSAND for describing complex sand mechanical behaviors, and the smoothed particle hydrodynamics (SPH) method for dealing with large deformation. To show the high efficiency and accuracy of the proposed approach, a series of column collapses using discrete element method (DEM) and considering the influence of particle shapes (i.e. spherical shape (SS), tetrahedral shape (TS), and elongated shape (ES)) were adopted as benchmarks and simulated by the proposed method. The parameters of SIMSAND were calibrated from the results of DEM triaxial tests on the same samples. Compared with the results of DEM simulations and reference solutions derived by published collapse experiments, the runout distance and final height of specimens with different particle shapes simulated by SPH-SIMSAND were well characterized and incurred a lower computational cost. Comparisons showed that the novel SPH-SIMSAND approach is highly efficient and accurate for simulating collapse, and can be a useful numerical analytical tool for real scale engineering problems.