Figure 1. Limiting hazard posed by smoke layer in an atrium.
The design and testing requirements for smoke management systems for covered malls and atria are included in NFPA 92B1. While covered malls and atria may appear to be different spaces for the purpose of limiting the hazard posed by smoke, they pose similar problems. Both covered malls and atria have large undivided spaces, relatively small amounts of fuel, and often have tall ceiling heights.

About 15 years ago, the design basis for covered malls and atria changed from providing a number of ach to providing the exhaust necessary (if at all) to limiting the hazard posed by smoke. In the model building codes in the United States, the default design requirement stipulates that the smoke layer cannot descend within 6 ft of the highest walking level that is part of the exiting path2. This requirement is depicted in Figure 1. The current hazard-limiting design necessitates that the design process consider the size of the design fire and conduct calculations to determine the characteristics of the smoke management system3. The focus of most smoke management system designs consists of the exhaust and makeup air supply capacities and their arrangement. Because acceptance test procedures need to be consistent with the design basis, the test procedures also had to change.

This paper describes frequent errors in design approaches and procedures for acceptance tests for smoke management systems in covered malls and atria that have been related to the author.

Figure 2. Smoke exhaust rate requirement.

Design Pitfalls

As with any fire protection system, the design of a smoke management system needs to be commensurate with the potential hazard posed by a fire. Consequently, selection of the design fire(s) is the essential first step of the design. The selection of candidate design fires seeks to identify reasonable scenarios that provide the greatest demand on the smoke management system (i.e., they represent "worst case" situations). The demand may be expressed in terms of the smoke exhaust capacity required, temperature of operating equipment, or maintaining a particular level of visibility or other smoke condition.

Optimistic Design Fire

The design fire is specified in terms of the fuel packages and their location in the covered mall or atrium. One of the principal parameters used to describe fuel packages is their expected heat release rate. Experimental data on heat release rates for many fuel packages is available in several references4,5,6 (Figure 2).

In an attempt to reduce the required smoke exhaust capacity, it's very tempting for designers to suggest that the maximum heat release rate expected for a fire in an atrium will be very small and the atrium will be a wide-open, pristine area with very few combustibles. However, does such a small heat release rate consider the presence of holiday decorations or weekend trade shows? Designing the smoke management system to react only to a small fire could significantly limit the use of the space. Reportedly, an optimistic selection of the design fire prevented a new arena from hosting a major trade show given that the commodities associated with the trade show had heat release rates that were greater than the design fire.

Doubling the heat release rate of the design fire only results in a 25% increase in the smoke exhaust capacity required. Being pessimistic when selecting the design fire may greatly expand the flexibility of the facility and permit it to be utilized in a fashion consistent with the vision of the owner. Revising the smoke management system to expand its capabilities after construction can be cost-prohibitive.

Figure 3. Diagram of plugholing.

Does High Capacity Exhaust Produce Plugholing?

The smoke exhaust rate needed to arrest the descent of the smoke layer is indicated in Figure 3. For large clear heights, the required smoke exhaust can be very substantial.

When providing large exhaust capacities, the quantity of exhaust to be included needs to be determined. If a lesser number of high capacity fans are selected as a means to provide the total exhaust required, a check needs to be provided to assess whether the "strength" of the fans is likely to create a hole in the smoke layer, as indicated in Figure 4. This phenomenon is referred to as "plugholing," a phrase that originated in British research literature. This phenomenon is similar to water draining from the bathtub. As the tub is near empty, a gurgling noise can be heard coming from the drain as air mixes with the water.

In general, plugholing occurs more easily as the temperature and depth of the smoke layer becomes smaller. As a result of a hole being created in the smoke layer, the exhaust fans will be removing air from below the smoke layer as well as the smoke. If plugholing occurs, the full capacity of the fans is not being utilized to remove smoke alone, thereby decreasing the effectiveness of the exhaust fans, leading to a descent of the smoke layer.

How Is Makeup Air Provided?

In order for the exhaust fans to be effective, makeup air needs to be provided below the design smoke layer position. In terms of the scientific principles affecting the performance of the smoke management system, the makeup air can be provided by any means, including natural ventilation (opening exterior doors) and mechanical supply fans.

Applicable codes may require a particular proportion of the makeup air to be provided mechanically. In general, relying on the leakage of exterior building components is not sufficient to satisfy the amount of makeup air supply required. The stipulation of makeup air provided is that it be limited to a velocity of 200 ft/min or less at the location of the design fire7.

Just as importantly, recent research has indicated the need to distribute makeup air supply points around the perimeter of the atrium or covered mall and not to concentrate the points only on one side. Computer modeling of situations where the makeup air is asymmetrically provided has indicated that such a design may deflect the flame to one side, thereby appreciably increasing the amount of smoke production, which in turn results in a lower smoke layer position. For this particular example, the smoke layer depth changes from 20 ft to 39 ft solely due to the change in the makeup air supply arrangement.

Figure 4. Experiment with love seat, 3 MW at 360 sec. (Photo courtesy of www.fire.nist.gov/fire/fires/fires3.html.)

Acceptance Testing

Acceptance tests of smoke management systems in covered malls and atria are conducted using a variety of techniques, ranging from rational approaches to those that are arbitrary. The most important consideration in developing a plan for an acceptance test for a smoke management system is that the test should seek to confirm that the performance parameters of the system are being achieved.

Figure 5. Minimum exhaust capacity to cause plugholing.

Criteria For Acceptance Test

One of the most significant flaws with the performance of acceptance tests involves the failure to stipulate acceptance criteria prior to conducting the test. This is analogous to playing a game without first deciding if more points wins or loses. Imagine the ensuing debates on a golf course or football field if the rules weren't established prior to the game. Acceptance criteria should relate to the design basis of the system, e.g., exhaust rates and makeup air supply rates.

Visible Smoke Tests?

The next most significant issue involving acceptance tests of these systems for some installations is what appears to be an automatic requirement of using visible smoke. While the use of visible smoke can be valuable in some applications, in other cases it can be used for the wrong reasons, perhaps giving a false sense of security or an inappropriate view of performance. Consequently, a particular engineering purpose for such a test needs to be identified prior to conducting tests with visible smoke, at least to determine appropriate acceptance criteria, if not to establish a relevant protocol.

Sometimes, the visible smoke tests are rationalized in terms of needing to simulate conditions during a fire. However, this is the only type of fire protection system for which such a demand is made. Simulated fire tests are not required of other active fire protection systems such as sprinklers or detectors, nor are they required for passive fire protection systems such as firewalls.

Some acceptance tests involve "filling the space" with smoke from smoke bombs and seeing how long it takes to clear the area or to observe smoke movement patterns. In the case of a "time-to-clear" measurement, smoke management systems in covered malls and atria are not designed to clear the space of smoke. As such, why should they be tested in that manner?

In contrast, needing to observe smoke movement patterns with the operating smoke management system is an appropriate concern. However, does the entire space need to be filled in order to conduct that assessment? Usually, the smoke movement patterns of interest are located at specific points, (near doorways, store front openings, etc.). Smoke movement patterns near openings can be observed in much simpler ways than filling the entire space with smoke. Small smoke sources can be used to produce visible smoke near the opening in order to observe the direction of smoke flow near that opening. Lightweight paper or air speed and direction measurements can also be used to indicate direction of travel.

Visible smoke tests may also be mandated because of a lack of confidence in the calculations or to confirm the appropriateness of assumptions. This is a reasonable concern, given the relatively contemporary nature of the computational techniques and the limited validation efforts for the application of these methods in the innovative designs that are often seen in covered malls and atria.

If the design fire is located away from the enclosing walls, where the smoke plume can rise to the upper portion of the space without encountering significant obstructions such as balconies or where the horizontal cross-section atrium gets narrower with height, one of the axisymmetric plume equations (equations 6.2.1.1b or 6.2.1.1c in NFPA 92B) have probably been applied to establish the design requirements. These particular equations were developed over 30 years ago, and their origin traces back to the mid-1950s. Their predictive capability has been examined numerous times, including one application in an arena reported by Dillon8. As such, there's little need to conduct a visible smoke test to confirm their accuracy. Doing such would be similar to confirming friction loss calculations provided by the Hazen-Williams equation in fire sprinkler piping, which is unheard of.

If a visible smoke test is being conducted to assess whether the smoke management system is able to arrest the descent of the smoke layer for other situations, the test scenario should seek to capture the same smoke production and movement mechanisms as in an actual fire. As such, cold smoke from smoke bombs should be avoided, as the smoke does not have the buoyancy of hot smoke and thus will move quite differently. In addition, the smoke production characteristics of smoke bombs are opposite to those of typical fires, where the smoke production rate of the smoke bombs decreases with time.

Heating the smoke should be done with extreme caution. Use of heating sources should be just sufficient to provide enough buoyancy for the smoke to reach the ceiling. Further, test engineers should confirm (prior to the test) that the heat output of any heating sources is appreciably less than that needed to cause any damage to the facility. Controls and extinguishment means need to be available to terminate the test promptly if necessary. The composition of the smoke bombs is also known to be carcinogenic, so anyone exposed to the smoke should wear self-contained breathing apparatus.

Finally, the movement of visible smoke will be influenced by the combination of forces affecting air movement in buildings that are present during the time of the test. As such, the performance observed during the test also will be dependent on the conditions present. If the conditions change, the performance may also change. Unless the presence of such forces is acknowledged and taken into account, a false sense of security may be created as a result of a successful visible smoke test on the test day when the conditions served to aid the performance of the system. ES