Issue



Airflow modeling in cleanroom design


05/01/2005







A technique that allows improvement of cleanroom configuration without interfering with normal manufacturing processes

Andy Manning, PhD, Flomerics, Inc.

Cleanrooms come in all shapes and sizes from special purpose minienvironments to large, expansive production facilities. These involve a variety of application areas within a range of industries. Some are quite obvious, such as almost all stages of pharmaceutical research, development and production, or semiconductor manufacturing, but others are less obvious although they adopt very similar ventilation strategies for the control and containment of airborne contaminants. They may be designed to protect product or personnel. The level of cleanliness has traditionally been associated with the method of ventilation; yet, the method of ventilation does not guarantee the ventilation performance because the airflow is extremely complex.

The bulk of the costs (capital and run-time) associated with the operations of a clean environment within the pharmaceutical and health-related industries is dictated by the desire to manufacture a high-quality, uncontaminated product, protect the personnel involved in the processes, and satisfy regulatory bodies such as the FDA.

The mechanical ventilation design in turn plays an extremely important role in determining whether or not the entire system meets the pass/fail criteria. At present, the rules defining the best ventilation design practices are based on simplistic historical data that are often wrong. The leading corporations in these industries tend to have their own in-house mechanical expertise that scrutinizes any proposed design. They tend to look at the final set-up and further verify the original assumptions prior to final acceptance.

Why? Because the performance of a cleanroom is defined by a set of complex interactions between the airflow, sources of contamination and heat, position of the air terminals and exhausts as well as the objects occupying the space in question. Consequently, changes to any of the above elements will have an effect on the outcome, invalidating the built-in assumptions of the empir- ical configuration.

In the ideal situation it is necessary to:

• Predict the performance of your clean environment prior to construction or mock-up.

• Be able to make changes to the layout of the room knowing the exact effect of such variations.

• Use only enough air changes per hour (energy conservation) to satisfy the requirements.

• Achieve the above in the shortest possible time.

Most importantly of all, it is necessary to do this at the lowest cost.

The only scientific way of achieving all of the above conditions is by solving for the physics of the problem using numerical methods known as Computational Fluid Dynamics (CFD). In the mechanical ventilation design fraternity, they are commonly known as Airflow Modeling techniques.

To demonstrate the power of these methods to identify design features that should be addressed prior to completion of the design, we will visit a variety of application examples. The aim is to show that the complex flows are such that they cannot be predicted prior to build and test.

Click here to enlarge image

null

What are CFD techniques?

These are mathematical modeling techniques used to solve for the physics of the complex fluid flow and heat transfer that occur in any real situation. The physics of fluid flow, heat transfer and associated processes are contained in the well-established form of the Navier-Stokes equation:

Unfortunately, the only way of solving this rather harmless-looking set of mathematical notations for any real life situation is to use numerical methods of CFD. The nuclear and aerospace industries were the first to invest in the development of CFD software programs in the late 1960s. Alas, due to the complex nature of these codes and the need for large amounts of computer processing power, their use has been restricted to the R&D departments of the largest of corporations in the world. However, with the advent of cheap computer power and the tailoring of CFD software for the specific industrial sectors, it is now feasible to apply these techniques to help with fast-moving Design Engineering applications.

In 1988, Flomerics identified the need for such industry focus. As a result, the FLOVENT CFD program for airflow modeling was developed to address the needs of the mechanical engineers designing ventilation systems for built environments. This was further backed by the creation of a FLOVENT Modeling Services department with the sole remit of training and providing consultancy to the building services industry.

How is it done?

Airflow modeling solves the set of Navier Stokes equations by superimposing a grid of many tens or even hundreds of thousands of cells that describe the physical geometry of heat and contamination sources and air itself. Figures 1 and 2 show a typical research laboratory and the corresponding space discretization, subdividing the laboratory into tens or hundreds of thousands of cells.


Figure 1. Geometric model of a laboratory.
Click here to enlarge image

The simultaneous equations thus formed are solved iteratively for each one of these cells to produce a solution that satisfies the conservation laws for mass momentum and energy. As a result, we can then trace the flow in any part of the room, simultaneously coloring the air according to another parameter such as temperature.


Fgure 2. Superimposed grid of cells for calculation.
Click here to enlarge image

In the case of the above laboratory, imagine that a smoke tube is positioned inside the supply-air diffuser and behind the make-up air grill (see Fig. 3). As the air moves, the smoke changes color according to its air speed. The intention of the Total Air Diffuser used in this design is to generate low velocities from the supply-air diffuser, which sweep the laboratory air to the hood exhaust. Low velocities from the supply air are shown in Figure 3 by the blue air. As air speeds increase, the color changes from blue to green to yellow to red. The red condition represents a velocity of 0.5 m/s (the mean sash opening velocity for the design) or more. Indeed, the cool supply air from the diffuser does not impact on the open sash of the laboratory hood, but falls to the floor. The warm air from the make-up air grill, however, floats across in front of the hood, destroying containment. Figure 4 shows the disturbed flow in front of the laboratory hood by placing the smoke source on the faces of an imaginary box extending approximately one foot outside the sash opening into the laboratory. Instead of the potentially contaminated air being entrained into the hood, it is swept around the laboratory by the cross-flow from the make-up air grill.


Figure 3. Flow from Total Air Diffuser and transfer. Courtesy of the National Institutes of Health.

Click here to enlarge image

The sensitivity to small changes in design can be seen clearly when we move the supply-air diffuser to the other side of the hood. The two incoming air streams now combine (see Fig. 5) with the result that the air from outside the hood is almost entirely re-entrained into the laboratory hood (see Fig. 6).


Figure 4. Disturbed flow in front of hood. Courtesy of the National Institutes of Health.
Click here to enlarge image

Of course it is not just the configuration of the room that affects the airflow, but also the overall construction of the building. Specifically, the supply-air plenum and the return-air void and chases have a dramatic impact.


Figure 5. Flow with Total Air Diffuser moved. Courtesy of the National Institutes of Health.
Click here to enlarge image

There are three primary benefits of using airflow modeling:


Figure 6. Flow re-entrained in front of hood.
Courtesy of the National Institutes of Health.

Click here to enlarge image

• The ability to anticipate potential design flaws so that they can be remedied before the facility is constructed


Figure 7. Airflow over equipment track.
Click here to enlarge image

• The ability to quickly and efficiently reveal areas of opportunity for improved performance in operating cleanrooms


Figure 8. Air circulation in the interface region.
Click here to enlarge image

• The ability to model a variety of options for both planned and operating cleanrooms so that the most economical solutions can be pursued with a high degree of confidence in their validity

Case study examples

On the largest scale, the traditional method of manufacturing semiconductors was to use a ballroom approach containing large manufacturing systems for the semiconductors. The approach is to supply air to the ballroom via a plenum (or set of plenums) through a ventilated ceiling normally consisting of HEPA filters. The air is intended to flow unidirectionally downwards (often incorrectly called laminar flow-the air is almost certainly turbulent). The air then passes through a partially open floor into a floor plenum where it is returned through limited return-air chases to the air-handling unit(s). The distribution in these floor and ceiling plenums is itself a critical issue, but these examples will concentrate on the cleanrooms themselves.


Figure 9. Original bath configuration. Courtesy of SGS Thomson Microelectronics.

Click here to enlarge image

Whatever the solution, the performance can be predicted using airflow modeling. Figure 7 shows laminar airflow (LAF) in a ballroom except where it is disturbed (like the airflow in the paint booth) close to the equipment. Each of the five large pieces of equipment consists of three parts: the loader (that can be identified by the tall column), the stepper (that can be identified by the large white cube), and the track (with a white lower region and a black upper section). A closer look at the interface region (see Fig. 8) between the stepper and the track shows that the local equipment effects can create undesirable air circulation. In this case, this is where the robot arm waits with the wafer between processes.


Figure 10. Revised extract flow rate and sash length. Courtesy of SGS Thomson Microelectronics.
Click here to enlarge image

This type of local condition can be critical for the semiconductor industry in the ballroom, as it is likely to result in a loss of yield. Close to the equipment, the airflow can be critical to the air quality. This example, courtesy of SGS Thomson Microelectronics, shows the effect of changing the local extract flow rates combined with a change in sash length. Air is supplied to the room through HEPA filters mounted in the ceiling above the equipment. The intention is to provide downward flow that prevents the chemicals from the chemical baths becoming airborne and being deposited around the space. Air is extracted from the back of the bench above the work surface to locally remove any of the evaporated chemical, with the remainder of the air supplied being recirculated underneath the equipment to the gray area behind.


Figure 11. Geometric model of vial-filling room. Courtesy of Chiron Corporation.
Click here to enlarge image

Figure 9 shows the airflow over chemical baths. This original configuration carries concentrations from the chemical baths over the sill. The inset, with a transparent view, shows that the contaminated air can clearly be seen traveling through the gray area behind as the air is recirculated, and it can also be seen as it circulates behind the sash. The result is that the chemicals are deposited as a fine powder on equipment. By modifying the airflow rates extracted from behind the chemical baths, and adjusting the sash length, the amount of chemical deposition can be reduced to an unnoticeable level (see Fig. 10). The actual settings required to optimize the configuration can easily be experimented with when using airflow modeling. In fact, sensitivity analysis of a design to changes in associated parameters is an ideal use of this analysis approach.


Figure 12. Flow in the original configuration.
Courtesy of Chiron Corporation.

Click here to enlarge image

This type of flow control is extremely important in manufacturing in the pharmaceutical industry. The unintentional flow regimes shown in the examples so far can be a great risk to the products, which must not be contaminated by other substances even in the smallest quantities.

The vial-filling room (see Fig. 11), addresses the effect of HEPA filter layout and exhaust grill layout on minimizing upward flow in the room. The flow visualizations are colored according to vertical velocity. This is a measure of airflow uniformity (laminarity) and thus of better ventilation design. To provide the aseptic environment, the key process areas are ventilated by HEPA filters from above. Above the fill equipment and in the laminar flow hood, the air is drawn from the room through high-level returns. The air is then filtered and resupplied to the room through the HEPA filters. Remaining HEPA filters are supplied from the main plant, with the return air exhausted from the room at low level. Other processes (e.g., the autoclave and oven, located on the long wall nearest the viewer) are not, however, completely surrounded by HEPA filters and so are more vulnerable to nonuniformity of airflow.

The design team at Chiron upgraded the fill machine and considered the possibility of using the opportunity to enhance the aseptic performance of the room. The chosen method was to use airflow modeling to understand the current performance, introduce the new equipment, assessing its impact on the ventilation performance, and to refine the design to minimize and contain any areas of upward flow. The consensus of opinion was that such a design outcome would lessen the risk to the airflow’s laminarity around the vials.


Figure 13. Local exhaust under the LAF.
Courtesy of Chiron Corporation.

Click here to enlarge image

The case study documented here deals with the latter part of the optimization. The current configuration with the new fill machine can be seen to perform well within the main laminar flow areas, however there are some features of the flow that are considered less than optimal. Figure 12 shows the air from the main laminar flow panels. In this view, air from the laminar flow hood hits the floor, turns back upwards into the LAF itself, and flows across the room before flowing upwards in front of the autoclave and oven.

Less visible in this view is the upward flow behind the pillar adjacent to the fill machine or the upward flow to the high-level returns above the fill machine. Both are a potential source for turbulent diffusion against the downward flow.

The process of redesign is an iterative one that involves inspecting the features of the flow and using engineering judgment to modify the configuration to resolve the perceived problems. In this case, the high-level returns are removed and the HEPA filter layout in the open part of the room is redesigned. Removing the high-level returns, as expected, has reduced the amount of upward flow, while placing the HEPA filters closer to the autoclave and oven protects them. Selective placement adjacent to the fill line has also reduced the risk of nonlaminarity.

Also, one solution for reducing upward flow in the LAF is to provide local exhaust around the corner of the room below the LAF (see Fig. 13). This reduces the quantity of air moving off across the floor towards the fill line, thus reducing the upward flow there.

Conclusion

Facilities engineers have identified significantly improved configurations for many types of cleanrooms without interfering with the normal manufacturing processes. The opportunities this technique provides allow facilities operators and designers to use the technique beyond simply improving the cleanliness of the environment. This can be undertaken to establish whether new designs, or changes to the design and layout of the equipment, protective curtains and so on, will enable more efficient use of the facility.

Andy Manning received his B.Eng and PhD from the University of Nottingham. He joined Flomerics, Ltd. in 1994, and transferred to Flomerics, Inc. in 1997 as the Technical Lead for FLOVENT operations. He is currently the director of thermal engineering, with direct responsibility for customer and consulting services for all Flomerics thermal software in the United States. He can be reached at andy.manning@flomerics.com.