University of Kentucky Cluster Fluid Dynamics (CFD) Group

COMPUTATIONAL FLUID DYNAMICS (CFD) DESIGN OF A HAND-DRYING DEVICE

Justin T. Brown*, George P. Huang*, Ralph Sias**, Stan Diniz**, William C. Mers Kelly***, Chuck Whatley***, and Greg R. Furnish***
*Department of Mechanical Engineering
University of Kentucky, Lexington, KY 40506
**Intecon Systems
Carlsbad, CA 92008
***MedVenture Technology Corporation
Louisville, KY 40217

Key words: Heat transfer coefficient, Hand dryer, CFD, Simulation

    Clean rooms around the world have been utilizing natural rubber and synthetic latex gloves as a means to prevent particle contamination in clean room processes and environments. The materials used in gloves degrade over time thus significantly reducing their effectiveness. The manufacturers currently minimize glove degradation by frequent glove changes during the workday. This process is costly to the manufacturer both in glove usage and worker downtime, because glove changes must take place outside of the clean room. The used gloves also create a significant amount of waste for our environment.
    Intecon Systems, Inc. has developed Binary Ionization Technology (BIT) to clean gloves inside the clean room and is using MedVenture Technologies for the industrial design of a BIT based glove cleaning machine.  The machine, with its short cycle times, allows gloves to be maintained at high levels of cleanliness and has the added benefit of allowing the reuse of gloves in some occasions where they might otherwise be discarded.  Binary Ionization is Intecon’s patent pending technology and is a process of passing a proprietary cleaning solution through an ionized gas prior to applying the solution to the object being cleaned (gloves, etc.).  In the machine, the worker’s gloves are subjected to a BIT enhanced spray, then are further cleaned, and dried by a balanced ionized air stream.  Particles released from the surface of the glove do not escape into the clean room due to the industrial design.

       Conventionally building several prototypes and testing each one with the intent to resolve issues through a trial-and-error process would achieve the design of such a device. The current paper presents the comparison between the simulation of a working prototype and a proposed prototype using commercial CFD software – Fluent [1]. The CFD analysis investigated the flow within the chamber of the device yielding values required to compare the benchmark to the proposed prototype. Comparisons between the two designs revealed significant differences in the heat transfer coefficient on the hand as well as the mass flow rate escaping the chamber through the hand opening.

    By looking closely at the flow, structure, and heat transfer coefficient of each design we can compare the two geometries without the cost of building the second prototype. It is important to be able to find out early in the design process which geometry is more effective because significant changes to the design require building new prototypes which are time consuming and costly.

    The original prototype, as shown in Figure 1-a, consists of a box with eight jet inlets on the top with spacing shown in Figure 2, one pressure outlet on the bottom, and two openings on the front wall employed to insert the hands into the device. The prototype was chosen for its simplicity to test this method of cleaning gloves. Although it was found to perform the task sufficiently, its crude shape and small openings left much room for optimization. A second design, as shown in Figure 1-b, was proposed that consisted of a more marketable cosmetic shape, one large opening for the hands, and fourteen jets on the top, with spacing shown in Figure 2. Before building another prototype and testing it for comparison to the original, a model of each geometry was created and simulated in Fluent [1]. Then a comparison between the two was established based on criteria including the heat transfer coefficient on the hands and the amount of particles escaping the openings. Once the comparison was made, adjustments to the body of the second geometry were incorporated to optimize its performance.

    The current paper presents several computational studies of the geometries in question using Fluent. Our objective is to compare the performance of a new geometry to the original prototype. The formulation of each CFD simulation and its complex 3-D grid generation are described in Section 2. Analysis of the computational results is given in Section 3 for both geometries. The results serve to provide a trend for the new geometry that can be used to define the geometry's performance compared to that of the original working prototype. Finally, conclusions of the study are given in Section 4.

                                

GRID GENERATION AND BOUNDARY CONDITIONS
    A major challenge in calculating the flow inside the hand-drying device is providing an adequate description of the geometry. Because of the complex geometry of the device, control over the grid is limited making it difficult to reduce the size of mesh without losing accuracy in the results. Also, the grid size is limited by the computer memory available. This forces us to rely on trends for the comparison instead of using the so-called “grid-independent solutions”. The definition of the complex geometry of the device and its grid generation were done using Gambit 1.0 [2], while the flow structure was solved using Fluent 5.0 [1].

Geometry Definitions
    The geometry for each device were generated as solid models using Solid Works and transferred to Gambit 1.0 using IGES files created directly from those solid models. Although Fluent claimed that Gambit was able to manipulate IGES files to form a virtual geometrical environment, we found that Gambit 1.0 is not able to process the current IGES files without the need for some additional clean-up.  For example, if two lines were connected at a single vertex in the solid model and the IGES file was transferred to Gambit, the software would create two vertices, instead of one, at the connection leaving the two edges unconnected. Although these two vertices consisted of identical coordinate values, Gambit was unable to connect them. This facilitated many problems in the generation of faces and volumes and forced the user to reconstruct the entire geometry satisfying the surface and volume definitions of Gambit 1.0.

    Many modifications must be performed during the geometry-definition process. The modifications include the elimination of any external features, since they are unimportant to the flow structure inside the chamber, and reconnection of all important edges for the generation of surfaces and volumes needed. Also, the hand geometry was measured from a human hand and inserted directly into Gambit using those dimensions to create cylinders for the fingers and wrist and a box for the palm. Because conduction within the hand was not a part of our analysis, we were able to generate two volumes, one for the chamber and one for the hand, and subtract the hand volume from the chamber volume to create one volume representing the air inside the chamber. The final geometries represent the exact shapes of the inner shells of the chambers, as shown in Figure 1.

    The volume grid was generated with volume grids in adjacent zones communicating with each other through their interfaces. The continuous mesh approach allows the flexibility to generate different size grids in different zones. For our analysis, accuracy near the hands, openings, and jet inlets deemed more important than other areas of the chamber thus smaller elements were used in these areas while larger elements were used throughout the rest of the chamber. In doing so, a reduction in the time required for the solution was implemented. Another big challenge presented itself when attempting to generate comparable grids for the two geometries because of significant differences in their shape and size. Because the only continuities between the two geometries were the hand and diameter of jet inlets, an identical mesh size was established for the hands' surfaces and jet inlet volumes. Then, edges along the walls of the chamber were meshed so that the chamber volumes would contain a similar amount of elements when Gambit generated their volume meshes.

Scaling Grids
    The complexity of the two geometries hindered the ability to sufficiently control the grid size for an accurate solution. However, a trend was discovered that increasing the number of grid elements yields a higher and seemingly more accurate solution for the heat transfer coefficient. With limited computer resources a sufficiently dense grid could not be generated, thus driving the analysis towards trends to compare the two geometries. To ensure accuracy in this comparison, a trend was developed by increasing the number of elements for each geometry. The scaling of each geometry's grid was governed by one rule, the surface and volume mesh sizes, of the hand and jet inlets respectively, must be decreased equally for both geometries. Therefore, a trend ranging from about 150 thousand to roughly 3 million volume elements was constructed. The magnitude of change from the first grid mesh to the final grid mesh is illustrated in Figure 3. Results of the comparison of the two trends are given in Section 3.

Boundary Conditions
    In the computation, the, top, bottom, and side surfaces were treated as adiabatic walls. The jet inlets located on the top of the chamber were considered to be constant velocity inlets (80m/s) with room temperature. The velocity was obtained from the prototype design.  Since the redesigned dryer contains more jet inlets than the prototype and the velocity remains the same, it has a larger mass flow. This increase in flow rate is justified by the type of pump intended for use with the new design. The hand and wrist were considered walls with a temperature 40K above the room temperature.  The pressure of the front opening and the exit holes of the chamber were assumed to be 1 atm and -375 Pa gage pressure, respectively.  The gage pressure at the exit was based on the measurement of the prototype design. The computation used the standard two-equation k-e turbulence model in conjunction with the standard wall functions.  The turbulence intensity at the inlet was assumed to be 2%.

            

    All computational results were obtained with the first order upwind schemes under the unstructured grid framework. Convergence was assumed when the residual of the continuity equation dropped more than 3 orders of magnitude (a default value recommended by Fluent). This usually leads to a reduction in the residual of the momentum equations of 4 orders of magnitude. Several computations were performed for two different geometries - one with a box shaped chamber, 8 jet inlets, and two small openings for hand penetration ( the original prototype) and the other with a more cosmetically shaped body, 14 jet inlets, and one large opening (proposed prototype). The predicted path lines for the two geometries were presented in Figure 4. Notice that with the original prototype, air inside the chamber circulates before exiting the chamber. This increases the possibility of particles being redistributed on the hand nullifying the purpose of the device. However, with the aerodynamic shape of the new design's bottom wall, the air is forced to escape through the exit minimizing this possibility. Although more air is force through the outlet of the chamber, there is also air forced towards the opening which may allow more particles to be re-deposited into the room.

         

    The performance of the hand dryer is driven mainly by how fast and evenly the hand will dry. This can be represented by the over all hear transfer coefficient defined as:

where Q is the total heat transfer rate, A is the contact surface area of the hand and Tj and Tw  are the temperature if the jet and hand, respectively. Because of the extreme differences between the two geometries, controlling the grid proved to be very difficult. To insure accuracy of the solutions, several cases were generated for each design by increasing the number of cells in the grid from case to case. Once the solutions converged, the heat transfer coefficient was calculated for each case and plotted against the number of cells used. From the plots generated, shown in Figure 5, a trend is revealed that the heat transfer coefficient increases with the number cells used.

    For the finest run we made (3 million points), it took about 2 Giga bytes of memory and 3 days to get a solution on the University of Kentucky HP N-4000 mainframe computer.  Due to the complexity of the problem, it appears that the so-called grid independent solution is impossible to achieve.  Fortunately, the general trends of the two solutions were a very similar as the grid is refined.  An underlying assumption is thus made that this trend will prevail as the grid-independent solutions are reached. The conclusion will then be drawn from the comparison of the solutions based on intermediate grids.

    From Figure 5, it suggests that the new design will increase the heat transfer coefficient by almost two times. This increase in heat transfer is driven by the addition airflow across the hands from the increased number of jet inlets. Another concern with drying the hand is to be sure that the entire hand dries evenly. The contour plot in Figure 6 illustrates a comparison of the local heat transfer coefficients on the hands between the two designs. The proposed prototype, Figure 6-b, depicts a significant improvement in the local as well as the overall distribution of heat transfer applied to the hand.

Particles Escaping through the Opening
    Another stipulation for the functionality of the device is the amount of particles escaping the chamber through the openings. Since the device is to be used in "clean room" environments, re-depositing particles into the room is not desirable to the manufacturing process. Although it is impossible to eliminate the possibility of particles escaping through the opening without using a sealed chamber, the number of particles should be kept to a minimum. From the solutions for each design a total mass flow rates at the openings were obtained using Fluent. Figure 7, shows the velocity contour plots at the openings of the two designs. The contour plot shown in Figure 7-b illustrates high velocities near the top, outer region of the opening. Although the contour plot only displays magnitude, by viewing the velocity vectors, in Figure 8, one can see that particles are escaping the new design. This is a downside to having the larger opening since mass flow exiting the chamber is a major concern. Our estimation based on the solution extracted from Fluent showed that the mass outflow for the new design is one order of magnitude larger than that of the original prototype design. This is a drawback of the new design that needs to be improved in future investigation.

CONCLUSIONS
    Two designs of a hand dryer developed by Intecon and Medventure to clean rubber or latex gloves were simulated using the commercial CFD software package - Fluent. The study serves to define differences between the two designs based on the heat transfer from the hand and the amount of particles escaping the openings of the chamber.

    The computations reveal that the new design of the device will improve the drying process by increasing both the local and overall heat transfer coefficients on the hand. However, the study also reveal that the new design will give rise to high particle escape rate as compared with the original prototype design.