Large Eddy Simulation of Flow-Field and Micro-Particle Deposition in an Idealized Mouth-Throat

The study concerns the simulation of the flow field and particle deposition in the upper human respiratory system. The numerical simulations are conducted using large eddy simulation (LES) and a Lagrangian particle tracking technique. To validate the numerical methods, the average gas velocity at the centerline and at different cross-sections in a constricted tube is compared with the experimental data [1] and numerical data from Reynolds-averaged Navier–Stokes (RANS) equations coupled with low Reynolds number (LRN) κ-ω model [2] and LRN shear-stress transport (SST) κ-ω model [3] from literature. Through the comparison, it is concluded that the present methodology is suitable to simulate the laminar–transitional–turbulent flow. In addition, the particle deposition efficiency in the idealized mouth-throat model is compared with data from the literature [2]. It demonstrates that the methodology can be used to predict particle deposition in the human oral airway. The main properties of the flow field in the idealized geometry are captured as other numerical simulation [2], but a small recirculation zone was observed in the posterior side of pharynx and laryngeal jet is observed to impinge on the anterior side of trachea wall in the present numerical result. The secondary vortices are attributed to the laryngeal jet profile. Introduction Aerosol particle deposition of therapeutic agents remains an essential tool in current treatment methods for asthma and other lung diseases. The advantage of pulmonary drug delivery through inhalation is that it offers topical treatment of specific lung conditions while limiting the whole-body effects [4]. The extra-thoracic region, including the nasal and oral passages, pharynx, and larynx, build the entrance to the human respiratory tract [5]. The therapeutic aerosol particles, drugs can be directly transported via the oral airway to the lung. Aerosol particles deposition in this region has important implications in drug delivery efficiency [5]. So, it is very important to study the airflow structures and particle transport for filtering effects in the oral airway [2]. One challenge for numerical simulation in this region is its geometric complexity [2]. There are a lot of previous numerical simulations that have been done on the idealized mouth-throat models [3]. According to dimensions of a human cast reported by Cheng et al. [6], an idealized oral airway model including mouth cavity, pharynx, larynx and trachea was built by Kleinstreuer et al. [7]. A series of research work [2,7-10] was done on this geometry model. For instance, the flow field and micro-particle transport in the idealized mouth-throat were simulated by Zhang et al. [2]. The numerical results in the simplified model exhibit the main features of laminartransitional-turbulent particle suspension flow in the human oral airways. Moreover, it shows that the turbulence may enhance the particle deposition in the trachea near the larynx, in particular, for the particle deposition of smaller particles. In 2004, the airflow and transport of nano-particles were analyzed by Zhang and Kleinstreuer [8] for both unsteady and steady inspiration flow rates in a upper airway including the oral airway. It is observed that the transient effects of flow field in oral airway mainly appear during the decelerating phase of inspiration circle. Another idealized mouth-throat was built by Stapleton et al. [11]. It was constructed on the information from literature, computed tomography (CT) scans, magnetic resonance imaging (MRI) scans, and direct observation of living subjects. Many numerical simulations [11-16] concern the mouth-throat model. Recently, numerical simulations using the realistic geometry are performed. The deposition pattern in a CT-based realistic airway configuration was analyzed by Jayarajua et al. [3]. The numerical results of flow show that the laminar to turbulent transition, especially at low flow rates, is sensitive to the complexity of the airway model. In fact, flow transition is seen soon after the glottis region for a low flow rate of 15 L/min, but which are not reported in the simplified geometry simulations [3]. Another challenge is that the flow ranges from laminar, transitional and turbulent within the respiratory system, which requires the method not only to capture laminar flows, but also transitional and turbulent flow structures [17]. Most of previous work used the RANS method to simulate the flow field. In 2001, Stapleton et al. [11] adopted the κ-ε model to simulate the flow field using a Lagrangian tracking method coupled with the eddy interaction model to simulate the fluid-particle in a simplified mouth-throat model. It is observed that the model *Corresponding author: [email protected] ILASS – Europe 2010 Large Eddy Simulation of Flow-Field and Micro-Particle Deposition in an Idealized Mouth-Throat 2 failed to predict the flow field in the relatively high flow rate and it is not suitable for the accurate prediction of particle deposition [11]. Four different turbulent models such as LRN κ-ε model, renormalization group (RNG) κ-ε model [18], which uses renormalization group method to account for the effects of smaller scales’ motion (http://www.cfd-online.com/Wiki/RNG_k-epsilon_model), Menter’s κ-ω model [18], which is suitable for high Reynolds number flows and LRN κ-ω, were used by Zhang and Kleinstreuer [18] to simulate the internal flow field in two different test conduits with local constrictions. The LRN κ-ω model is identified to be more suitable to simulate the laminar-transitional-turbulent flow in the constricted tube [18]. The LRN κ-ω model was widely used in numerical simulation of the flow field in the respiratory system [2,7-10]. In addition, LRN SST κ-ω model is proved to predict the transitional flow accurately [3]. The RANS model is suitable for fully developed turbulence, but it may be inappropriate for particle transport in the region with complex flow such as upper respiratory tract [19]. Recently, the prediction of particle deposition has been implemented more and more with large eddy simulation (LES). Luo et al. [20] used LES to simulate the flow in a single asymmetric bifurcation model and a constricted tube. It was demonstrated that LES predicts the transitional flow in the constricted tube better than the LRN κ-ε model. Jin et al. [14] simulated the flow and micro-particle deposition in a three-dimensional geometric model of human upper respiratory tract. It is found that turbulent dispersion plays an important role in the particle deposition for the particles with small Stokes number. It is observed that particles with the diameter of 1 μm not only deposits on the opposite wall but also on the side wall. Jayarajua et al. [19] simulated the fluid flow in a human mouth–throat model under normal breathing condition (30 L/min) alternatively using RANS κω (without near-wall corrections), detached eddy simulation (DES) and LES methods. By comparison with existing experimental data in situations below 5 μm and bigger particles, it is found [19] that for the medication aerosols inhaled at a steady flow rate of 30 L/min, LES and DES provide more accurate results than the RANS κ-ω model in predicting particle deposition. Also, both the LRN κ-ω and LRN SST κ-ω model have been evaluated for the flow field in the constricted tube through comparison with experimental data and other RANS models, and they are frequently used in the numerical simulation of particles transport and deposition in the respiratory system as mentioned above. However, no comparison has been made for the numerical results using LRN κ-ω, LRN SST κ-ω and LES [17] in the constricted tube. In this paper, the validation of the methodology with LES will be performed through the comparison of flow field in the constricted tube with RANS LRN κ-ω model and RANS LRN SST κ-ω model. Moreover, the flow field and particle deposition in a circular simplified mouth-throat were conducted using LES and a Lagrangian particle tracking method. Geometrical Models Two different geometries are used in the numerical simulations. The first one is an axi-symmetric constricted tube with an area reduction of 75 %, which is depicted by a cosine function [18] as: