Verification of numerical results with experimental data is an important aspect of any in silico study. In the case of the upper respiratory system, the air flow is often turbulent, which highlights the importance of validating an accurate turbulence model for numerical simulations. Patient specific CT based upper airway models were used for computational fluid dynamics (CFD) analyses of the upper respiratory system and the results were compared with the corresponding experimental results. Detailed CFD simulations were conducted using the STAR-CCM+ software to investigate the most appropriate numerical approach in accurately predicting flow characteristics in the upper respiratory system. Large Eddy Simulations (LES) and Reynolds-Averaged Navier-Stokes (RANS) equations with k-ε, and k-ω turbulence models were investigated.
The experiments include simulating inspiratory-expiratory flow with particle injection at the intake. A stereolithographic (SL) system (3-D system Projet 6000HD), with a resolution of 0.001–0.002 inches per inch of part and VisiJet SL Clear material, was used for fabricating the experimental model. The outlet of the model was connected to a manifold, with subsequent connection to a piston-cylinder system where a computer-controlled motor was used to simulate the normal breathing flow conditions. Investigations of flow characteristics within the upper airway were performed with a 2-D µPIV system from Intelligent Laser Applications (ILA for micro particle image velocimetry) which includes a high power green LED light source with an effective area of 100×100 mm, and a pulsing system (LPS controller). Matlab software was used for the post processing of PIV images.
The LES results displayed more detailed transient flow characteristics than the RANS results for both turbulence models. At the early time steps, the numerical results of the average velocity from all three methods were nearly identical. However, further downstream, where obstructions and strong velocity gradients exist, results differ with a larger velocity gradient near the wall for the LES simulation. Comparing the numerical and experimental results, due to seeding limitations, the experimental results did not display detailed low speed flow characteristics and thus, the shear stress and turbulence quantities were less than the corresponding CFD results. Further experiments are currently in progress to improve the experimental results and to better assess the transient numerical and experimental results.