The Motion solver offers several linear solver types to accommodate the diverse needs of multibody dynamics simulations. These linear solvers are designed to efficiently handle the complex interactions and computations required in these analyses. When the option is selected in the analysis settings, the Motion solver automatically defaults to using the Sparse Solver.
This solver leverages the Intel MKL library, a commercial library known for its computational efficiency. The Sparse solver uses a Compressed Sparse Row (CSR) format for matrix representation. Its performance is consistent across different computing platforms, making it a versatile choice for various simulation requirements.
The Frontal solver employs the nested dissection method to eliminate the intermediate variables of the system equations. Because this method does not save the intermediate values, it requires less memory than the Sparse solver option.
Expanding upon the Frontal solver, the Super solver saves and reuses LU decomposition values as long as NR iteration converges. This approach is efficient for simulations involving slowly moving contacts or small deformations. The design of the Super solver allows for assigning each flexible body to a different core in HPC implementations, enhancing performance for simulations with multiple flexible bodies.
The Preconditioned Conjugate Gradient (PCG) solver is an iterative solver tailored for small deformation models. While it is generally slower than traditional linear solvers for large deformation or shell element problems, the PCG solver is effective in scenarios where a suitable preconditioner can be identified during preprocessing. The performance of the PCG solver varies with the simulation's nature, necessitating extra time to find an effective preconditioner. PCG only supports FE Nodal Flex bodies. If EasyFlex bodies are included or modeled solely as rigid bodies, PCG cannot be used. The PCG solver is generally optimal for large-scale problems, approximately greater than 200,000 DOFs.