Table 5:
Comparison of the performance of different UQ methods. Each of the four UQ methods compared (Monte Carlo, multilevel, MLMF with 3D, 1D, and 0D models, and MLMF with only 3D and 1D models) requires differing costs to achieve a nCI value of 0.01. For four representative outlet flow, outlet pressure, time-averaged model pressure, and TAWSS QoIs, the cost of each method (in units of the equivalent number of 3D fine simulations) is shown. The same representative QoI is used for both healthy and diseased models. These QoIs are spatially and temporally averaged over the outlet face, throughout regions of the model volume, or over regions of the model surface for the flow and pressure, model pressure, and model TAWSS QoIs respectively.
| Healthy Model | Diseased Model | Healthy Model | Diseased Model | ||
|---|---|---|---|---|---|
| QoI and Method | Effective Cost | Effective Cost | QoI and Method | Effective Cost | Effective Cost |
| MC | 9 758.0 | 9 471.0 | MC | 10 656.0 | 10 619.0 |
| MLMC | 1 293.6 | 2 593.8 | MLMC | 435.1 | 1 095.5 |
| MLMF (3D-1D) | 77.9 | 468.3 | MLMF (3D-1D) | 41.0 | 49.4 |
| MLMF (3D-1D-0D) | 66.5 | 440.2 | MLMF (3D-1D-0D) | 39.5 | 43.7 |
| MC | 19 340.0 | 18 374.0 | MC | 20 467.0 | 20 587.0 |
| MLMC | 2 915.4 | 4 588.1 | MLMC | 948.1 | 2 480.6 |
| MLMF (3D-1D) | 82.2 | 282.2 | MLMF (3D-1D) | 46.6 | 65.1 |
| MLMF (3D-1D-0D) | 107.0 | 282.2 | MLMF (3D-1D-0D) | 40.4 | 44.7 |
| MC | 19 598.0 | 19 050.0 | MC | 20 213.0 | 20 330.0 |
| MLMC | 3 016.3 | 4 430.4 | MLMC | 934.1 | 2 445.8 |
| MLMF (3D-1D) | 119.4 | 321.9 | MLMF (3D-1D) | 45.1 | 65.0 |
| MLMF (3D-1D-0D) | 210.5 | 369.2 | MLMF (3D-1D-0D) | 40.4 | 46.4 |
| MC | 23 94.0 | 22 581.0 | MC | 21 748.0 | 16 391.0 |
| MLMC | 5 578.7 | 4 695.4 | MLMC | 2 066.7 | 2 391.4 |
| MLMF (3D-1D) | 1 151.4 | 1 922.4 | MLMF (3D-1D) | 901.9 | 1 303.8 |
| MLMF (3D-1D-0D) | 581.2 | 847.0 | MLMF (3D-1D-0D) | 290.6 | 269.7 |
| (b) Coronary Model | |||||