Computational modeling of peri-aneurysmal hemodynamics is typically carried out with commercial software without knowledge of the sensitivity of the model to variation in input values. For three aneurysm models, we carried out a formal sensitivity analysis and optimization strategy focused on variation in timestep duration and model residual error values and their impact on hemodynamic outputs. We examined the solution sensitivity to timestep sizes of 10-3 s, 10-4 s, and 10-5 s while using model residual error values of 10-4, 10-5, and 10-6 using ANSYS Fluent to observe compounding errors and to optimize solver settings for computational efficiency while preserving solution accuracy. Simulations were compared qualitatively and quantitatively against the most rigorous combination of timestep and residual parameters, 10-5 s and 10-6, respectively. A case using 10-4 s timesteps, with 10-5 residual errors proved to be a converged solution for all three models with mean velocity and WSS difference RMS errors less than <1% compared with baseline, and was computationally efficient with a simulation time of 62h per cardiac cycle compared to 392h for baseline for the model with the most complex flow simulation. The worst case of our analysis, using 10-3 s timesteps and 10-4 residual errors, was still able to predict the dominant vortex in the aneurysm, but its velocity and WSS RMS errors reached 20%. Even with an appealing simulation time of 11h per cycle for the model with the most complex flow, the worst case analysis solution exhibited compounding errors from large timesteps and residual errors. To resolve time-dependent flow characteristics, CFD simulations of cerebral aneurysms require sufficiently small timestep size and residual error. Simulations with both insufficient timestep and residual resolution are vulnerable to compounding errors.
ASJC Scopus subject areas
- Orthopedics and Sports Medicine
- Biomedical Engineering