1、Author Andrew E. Byl is a graduate student in the Department of Mechanical and Industrial Engineering, Montana State University, Bozeman, Montana. Authors Kevin L. Amende and Erick L. Johnson are professors in the Department of Mechanical and Industrial Engineering, Montana State University, Bozeman
2、, Montana Using Computational Fluid Dynamics to Characterize Airflow Through an Air Handling Unit Andrew E. Byl Kevin L. Amende, PE Erick L. Johnson, PhD Student Member ASHRAE Associate Member ASHRAE ABSTRACT HVAC equipment manufacturers spend a considerable amount of time and effort updating existi
3、ng product lines in order to meet ever-increasing performance standards. Traditional approaches consisting of several prototype iterations being built and experimentally tested are time consuming and costly. Through advancements in computer technologies within the last decade, computational fluid dy
4、namics (CFD) has become an economical solution allowing HVAC equipment designers to numerically model prototypes and reduce the time required to optimize a given design and identify any potential failures points. Airflow uniformity is considered an important consideration for air handling unit (AHU)
5、 manufacturers as it affects the performance of the overall system. Plenum fans inherently produce a rotational airflow pattern, which, if not mitigated appropriately, can create a highly, non-uniform airflow that enters a heat exchanger located downstream. This can lead to lower heat transfer rates
6、 and premature heat exchanger failure. While CFD offers the ability to visualize and characterize the airflow through an AHU system, it is often used to solely model individual components such as fans or heat exchangers without analyzing it in a real world situation. This paper presents the CFD mode
7、ls used to characterize the airflow uniformity within an AHU in order to aid in understanding heat exchanger performance. INTRODUCTION Heating, ventilation, and air conditioning (HVAC) systems around the world have provided a comfortable indoor living/working environment for billions of people and t
8、he requisite temperature control required for many biomedical and manufacturing processes. In 2009, 47% of energy consumption in residential buildings was used toward heating and air conditioning (Residential Energy Consumption Survey (RECS) 2009) while in 2014, 41% of the total U.S. energy consumpt
9、ion was consumed in residential and commercial buildings (U.S. Energy Information Administration 2015). Today the need for highly efficient HVAC systems is ever growing, requiring designers to continuously update products in order to balance this demand with the cost of energy. In the past, the desi
10、gn and optimization of these systems had solely been an iterative process with physical systems, which is time consuming and costly. More so, many AHU manufacturers lack the ability to visually characterize the airflow through such a system, as with particle image velocimetry (PIV) that uses laser r
11、eflections off in-situ particles to map airflow velocity. Even with the use of PIV techniques, a new physical prototype would still be required for each substantial design change. Computational fluid dynamics can be a powerful alternative for this application, allowing multiple iterations of numeric
12、al models to be used and in lieu of a physical unit. HVAC industry researchers have used CFD analysis in the past to model the performance of hardware such as blower fans, heating and cooling coils, filters, and even ductwork (Patel and Patel 2013, Bhutta, et al. 2012). However, it appears that CFD
13、use within the industry is still in its nascent stages and has primarily been used to model isolated components instead of as part of an entire system. One aspect AHU manufacturers have begun to consider when trying to increase efficiency is the airflow uniformity (Bhutta, et al. 2012, TJoen, De Pae
14、pe and Vanhee 2006). Flow uniformity at a particular point can be defined as the deviation of the local velocity from the mean plane velocity. A non-uniform flow is generally created from the centrifugal effects and turbulence generation of the blower unit (Jairazbhoy et al 2009). Research has shown
15、 that a non-uniform airflow across a heat exchanger can cause the global heat transfer coefficient to drop as much as 8.2% (TJoen, De Paepe and Vanhee 2006). This loss of performance in turn requires the blower unit to run longer, increasing the maintenance cost and decreasing the efficiency of the
16、whole system. CFD provides a way to quantify the flow uniformity at all locations in the AHU, as well as evaluate the effects it can have on the heat transfer performance of the heat exchanger, without the need to build a physical mockup. This study presents the CFD models used to characterize airfl
17、ow uniformity in an AHU and its impact on heat transfer. Data was collected in lab tests in order to tune the CFD models, where the airflow uniformity was then analyzed for two volumetric flow rates within a AHU. Experimental velocity data was then compared with simulated values in an AHU without an
18、d with a flow-straightening baffle. A new numerical model, consisting of only the AHU downstream of the blower, was created and a new inlet was specified using velocity data from the previous simulation. This allowed the simulation to model the downstream air characteristics at a greatly reduced com
19、putational cost. EXPERIMENTAL SETUP Three experimental tests took place within the psychrometric chamber located in the HVAC Laboratory at Montana State University. The chamber consists of two environmentally controlled rooms that allow for mechanical systems, such as an AHU, to operate at condition
20、s simulating combinations of indoor and outdoor environments. Straight Duct The first experiment consisted of obtaining the pressure drop across a single cooling coil installed in a straight duct, shown in Figure 1. A 0.381 m x 0.508 m x 1.829 m (15 in x 20 in x 72 in) straight duct section was util
21、ized as the upstream/inlet section, with a 0.584 m x 0.457 m x 0.914 m (23 in x 18 in x 36 in) section on the downstream side. The coil was mounted in the larger duct, at the interface between the upstream and downstream sections. Velocity data and the pressure drop across the coil was recorded for
22、13 flow rates ranging from 11.81 to 53.09 m3/min (417 to 1875 ft3/min) using a velometer (2% error) and two static pressure probes (5% error). Figure 1 Straight duct experiment used to obtain the pressure drop across the cooling coil Full AHU The second and third experiments were performed using an
23、AHU equipped with the same coil tested in the first experiment and a blower unit, shown in Figure 2. Velocity was recorded across a plane (see Plane 3 in Figure 4) at fan speeds of approximately 1000 RPM and 1800 RPM. Holes were drilled at the center edges of this plane, and velocity data was taken
24、with an anemometer at 0.0254 m (1 in) increments (see Lines 1 and 2 in Figure 4). The blower speed was controlled via a potentiometer and measured using a stroboscope. A heat exchanger downstream of the fan was not included in this experiment in order to focus on the downstream airflow characteristi
25、cs produced by the fan. Once the unobstructed experiment was completed, a tapered, L-bracket baffle design, shown in Figure 2(b), was added to the AHU and the experiment was repeated at both blower speeds. Figure 2 (a) AHU used in the experiments (b) Baffle design used in the experiments NUMERICAL S
26、ETUP Geometry To reproduce the experimental results, a 3D computer-aided drafting (CAD) program was used to create the straight duct and both full and simplified AHU geometries. Native CAD files used for manufacturing drawings often contain small gaps and features that make them unsuitable (or unnec
27、essary, e.g. small fasteners) for CFD simulations. Additionally, CFD simulations require the space the air occupies to be modeled, requiring new geometry files. CFD-appropriate geometry was created and imported into a commercially available CFD code. Straight Duct. Three separate regions were create
28、d for each of the inlet, coil, and outlet of the straight duct experiment. Instead of modeling the complex geometry of the coil, a rectangular, anisotropic porous region was created, with the porous inertial and viscous loss coefficients tuned to match the pressure drops from the experiment. Due to
29、the coil being located upstream of the blower in the AHU, the specific airflow characteristics induced by the coil are not relevant and the porous region provides an appropriate surrogate for the pressure drop across the coil at all velocities. The simulation was run at the same inlet velocities as
30、the experiment and the pressure drop across the coil was recorded. The pressure drop versus inlet velocity for the tuned values can be seen in Figure 5. Full AHU. The CFD models of the AHU consisted of the same components used within the experimental tests including the base AHU assembly, the blower
31、 unit, the coil upstream of the blower (again, approximated as a porous region), and the flexible ductwork. The AHU was split into four regions for the inlet, coil, blower, and outlet, with adjacent regions connected through stationary interfaces (Figure 3(a). As in the experiment, the rotation of t
32、he blower unit was the driving force of the airflow. This resulted in setting the inlet surface boundary condition as a stagnation inlet, which allows the necessary amount of mass to infiltrate the system, and the outlet surface boundary as a pressure outlet boundary condition. A rotation was applie
33、d to the blower region, causing it to rotate about its central axis at the same speeds measured in the experimental testing. The coil region used the loss coefficients found in the straight duct simulations. A second AHU model was created to incorporate the baffle design utilized in the third experi
34、ment of this project (Figure 3(b). Figure 3 (a) Full AHU Model and (b) AHU model with baffle Simplified AHU. The time required to run a CFD simulation for the model shown in Figure 3 varied between multiple days to over a week, to reach a fully developed flow and depending on the length of simulatio
35、n time desired. To reduce the computational time, a simplified AHU model was created by removing all the geometry located upstream of Plane 1 (Figure 4(b). Velocity data was recorded on three different planes from the full AHU model (Figure 4(a) over multiple time steps and the time-averages across
36、Plane 1 were used as the inlet velocity conditions in the new model. A heat exchanger region was added to assess heat transfer performance. This region was assumed to be identical to the cooling coil from the full AHU model, utilizing the same loss coefficients to model a pressure drop across the he
37、at exchanger. A porosity of 0.4 was used and affects the density evaluation in the momentum equations. NUMERICAL MODELS The Unsteady Reynolds Average Navier-Stokes (URANS) equations coupled with a k turbulence closure model were used to predict the turbulent airflow through the AHU. These models wer
38、e chosen due to their accuracy versus computational cost. The straight duct and AHU geometries utilized a polyhedral mesh. Prism layers were utilized in the blower region to increase the accuracy of the boundary layer around the fan-blade walls. The simulation was initially run multiple times using
39、a variety of mesh sizes and it was determined that a mesh independent solution existed for the AHU model. The inlet region used a base mesh size of 0.04 m (0.39 in), with target sizes of 0.0175 m (0.69 in) around the blower region and 0.02 m (0.79 in) for the coil and outlet downstream of the blower
40、. The coarser mesh size was chosen to reduce the computational cost in the regions where the flow would not have significant variations. The smaller base size in the blower region was chosen to more accurately capture the airflow near the fan while still allowing a smooth transition between other re
41、gions. the mesh consisted of between 550,000 to 750,000 cells for the full AHU models without and with the baffle included. To capture the rotational effects of the fan, time steps of 1.667e-4 s and 9.259e-5 s were used for the 1000 and 1800 RPM simulations, respectively. This is approximately 1 deg
42、ree of rotation per time step, with a time-step independnence study also having been performed. Figure 4 (a) Planes and lines where velocity data were recorded. (b) The Simplified AHU model RESULTS however, there was only a slight change in the lateral velocity components, and no correlated trend in
43、 either decreasing or increasing the variation in the z-component of velocity. Observing the z-component velocity profiles in Figure 8 it becomes more apparent that the baffle changed the overall flow structure by disrupting the vortex core. This signifies that while the baffle is successful at alig
44、ning the flow, it is not substantially impacting the turbulence or the variation in velocity-component magnitudes. Table 1 and 2 also show the flow uniformity for the simplified models with a heat exchanger region included in order to understand the effect of flow uniformity on heat transport. A hea
45、t transfer rate of 15.0 kW (51,182.12 Btu/hr) added energy into the simplified AHU models. The flow uniformity of both the 1000 and 1800 RPM models takes a slight hit when a heat exchanger is added (Table 1). Since the heat exchanger is treated as a porous region, the new obstruction creates a high
46、pressure area and slows down the overall flow, filtering a highly rotational flow into one that is more aligned as it exits this region. This high pressure region also shifts the flow structure upstream of it, which can be seen in Table 2. Comparison of the heat exchanger performance shows a decreas
47、e in heat transfer of 0.04% when the baffle is added to the flow path in the 1000 RPM fan speed models (16.5 kW and 16.5 kW (56,300 and 56,300 Btu/hr) respectively), and a decrease of 0.7% in the 1800 RPM models (17.0 kW and 16.9 kW (58,006.0 and 57,665 Btu/hr). The small (and negative) change in th
48、e amount of heat exchanged to the flow is insensitive to the large improvement in flow uniformity. This is also not captured in trends with the RMSE of the velocity components. While the addition of the baffle does improve flow uniformity, other factors need to be considered in order to properly cha
49、racterize effects of any baffle on flow uniformity, turbulence, and heat transfer. Table 1. Flow Uniformity Values for Plane 2 Without Heat Exchanger With Heat Exchanger Fan Speed, RPM No Baffle With Baffle No Baffle With Baffle 1000 55.9% 66.0% 56.0% 64.3% 1800 63.3% 69.0% 59.6% 67.8% Table 2. RRMSE Values for Plane 2 Without Heat Exchanger With Heat Exchanger Fan Speed, RPM Velocity Component No Baffle With Baffle No Baffle With Baffle 1000 X 64.0% 66.2% 70.1% 64.7% Y 70.4% 69.3% 70.0% 73.8% Z 48.1% 53.0% 38.6% 34.0% 1800 X 63.2% 64.0% 69.2% 64.7% Y 72.5% 70.0% 70.6% 77.4% Z 47
