| description abstract | Simulating large-scale problems are still challenging for discontinuous deformation analysis (DDA). To this end, this paper develops an efficient disk-based DDA (DDDA) model considering the efficiency in searching contact pairs and solving linear equations. First, an efficient contact search algorithm, namely the lattice search algorithm (LSA), is proposed, and efficiency tests of the LSA and direct search algorithm (DSA) demonstrate the high efficiency of the LSA. Second, three equations solvers, namely Jacobi iterative method (J), conjugate gradient method (CG), and preconditioned CG (PCG), are adopted to respectively solve the equations of the DDDA, and efficiency tests of these solvers show that the best solver is the PCG and the J is unsuited to solving the equations when using large penalty spring stiffness. Finally, a landslide simulation which includes 30,000 disks, 5,990 line segments, and 180,000 calculation steps is conducted, the result of which shows that: (1) up to 41.1 h, which is 99.2% of the time consumed in contact search using the DSA, is reduced by using the LSA; (2) up to 18.22 h, which is 69.8% of the time consumed in solving the equations using the CG, is reduced by using the PCG as the equation solver; (3) the time consumed in the simulation is 68.72 h when using the DSA to search contact pairs and using the CG to solve the equations; while the consumed time reduces to 9.4 h when using the LSA to search contact pairs and using the PCG to solve the equations, whose reduction proportion is 86.3%. The simulation indicates the large-scale computation capacity and further application in engineering with the DDDA. | |
