Issue link: https://www.e-digitaleditions.com/i/1475629
Inhalation August 2022 15 Model geometry preparation e fluid volume ("domain") to be simulated can be generated using the following approaches: • From a 3-D computer-aided design (CAD) model of a device or product—at a well-known, recorded state of development. e negative of the solid model can be used to create the fluid inside/around it for simulation. • Created manually within the CFD software. Care must be taken to ensure that it is representative of the real situation to be simulated. • From 3-D data from computed tomography or magnetic resonance imaging (MRI) scans. Manual/automated manipulation may be needed to smooth or patch image data into a usable closed volume for CFD. e mesh Once the fluid volume is created, it must be discret- ized into cells. is procedure has great impact on the simulation's accuracy and the computational time [30], so care must be taken. Guidelines are available about mesh style for airway models [30]. An auto- mated mesh generator within a CFD package or stand-alone package can create cells of hexahedral or tetrahedral shape, in either a uniform ("structured") or semi-random ("unstructured") pattern. At curved edges, hexahedral cells may have one face reduce in approach, typically large eddy simulation (LES) to resolve large turbulent flow structures. Better repre- sentation of the flow allows for greatest insight into underlying physics and subsequent development of new models and computational processes. LES modeling requires much more computational time. Reference 14 notes that "each calculation requires 15,000 to 35,000 CPU-hours (3-4 days of run time) depending on the grid size. is yields 0.16 seconds of a simulated pMDI spray." erefore, several hun- dred CPU cores were used, running in parallel. In selecting software to analyze problems relevant to specific products, the usual questions of cost and availability will be important, as well as consideration of the software's modeling capability. Study of aero- sols requires solvers for Lagrangian particle tracking that allow phase change and particle temperature variation. Internal flow studies require incorpora- tion of compressible flow solvers within the soft- ware. User-defined function may be a necessity, as employed in references 18 and 19, and as discussed in the section "Solver physics and sub-models" in this article. e following sections describe the typical CFD workflow. As with experimental or analytical pro- cesses, adherence to a proper procedure can help ensure confidence and traceability of CFD results and any decisions made on the basis of these.