Information identified as archived is provided for reference, research or recordkeeping purposes. It is not subject to the Government of Canada Web Standards and has not been altered or updated since it was archived. Please contact us to request a format other than those available.
Construction Technology Update No. 72, Dec. 2008
By G.D. Lougheed
This Update presents the results of research on smoke movement through HVAC systems and the effectiveness of duct smoke detectors. It provides guidance to practitioners and regulatory authorities in the context of North American code requirements for such detectors.
Requirements on the use of smoke detectors in HVAC ducts are included in most building codes, including the National Building Code of Canada (NBC).1 The intent of the requirements, based on a recommendation made by the National Board of Fire Underwriters in 1939 2, is that the HVAC system be shut down during a fire to minimize the circulation of smoke through the building by the HVAC fan system. In most cases, the detectors must be located in the supply air duct downstream of the fresh air inlets, filters and fans (see Figure 1). In some jurisdictions, detectors must be located in the return air duct as well. Installation requirements are provided in standards, including CAN/ULC-S524.3
Because there had been little or no research or data analysis to support this requirement, the Fire Detection Institute* undertook research on duct smoke detection to provide technical data to codes and standards committees and system designers. The research, conducted jointly by NRC-IRC and the University of Maryland (UMD), examined the use of duct smoke detectors both as a means of detecting fires or smoke within the HVAC system and as part of a building's smoke management system. UMD conducted small-scale experiments and modelling studies while NRC-IRC conducted full-scale experiments using its ten-storey test facility.
* The Fire Detection Institute has recently merged with the Fire Protection Research Foundation.
The research project addressed issues that had been raised regarding the need for duct smoke detectors and how they are installed and used. The key question was: Does duct smoke detection work and is it worth the added cost, considering the potential for false and nuisance alarms? The specific issues investigated were the comparative driving forces of the HVAC fans relative to those produced by the fire; how detection is affected by smoke dilution, smoke aging, type of HVAC filter, and stratified flow in the HVAC ducts; and the efficacy of sampling tubes used for duct detection.
Comparative Driving Forces
The shutdown of the HVAC system is intended to minimize smoke transport through the building by the HVAC fans. It does not eliminate smoke movement through building shafts (elevators, stairs and service) and ductwork as a result of pressure differences produced by the fire and ambient conditions (stack and wind effects). The research examined the issue of whether or not smoke movement created by the HVAC fans is significant relative to that resulting from the fire itself and other effects.
It was found that the HVAC-related pressure differences were generally larger than those stemming from other factors, including the fire itself. These greater pressure differences also led to higher flows and the distribution of smoke to floors where there was no fire. The results confirm that unless an active smoke management strategy is in place, the HVAC fans should indeed be shut down upon detection of a fire, as required by codes and standards. The extent of the advantage of shutting down the system depends on the specific characteristics of the building in question. This can be quantified using computational methods.4
Engineers and code officials have long been concerned that the concentration of smoke in the HVAC ducts might be too low for detectors to respond reliably to the fires they are generally expected to detect. To address this concern, NRC-IRC researchers conducted experiments using four different types of commercially available duct smoke detectors — ionization, photoelectric, sampling and multi-sensor. They compared the analog output* (signal) from each detector with the optical density of the smoke measured at the same location in the return air duct in the HVAC system. The measurements related to dilution and aging effects were made in a return air duct. However, the results would be the same for detectors located in the supply duct.
* The analog signal from the detector is sent to the alarm control unit, where the levels at which the alarm signal would sound are selected. Each detector used in the research had a different range of outputs. For comparison purposes, the outputs were converted to a percentage scale based on the maximum output for each detector.
As shown in Table 1, the output of the detectors was proportional to the optical density of the smoke. The analog output at which the detectors would produce a trouble signal or an alarm is set at the control unit and is typically less than the maximum output from the detectors. As indicated in Table 1, all the detectors would respond to smoke in the HVAC system once it reached concentrations that are comparable to those used as criteria for safe building evacuation.
|The measured optical densities can be used to estimate the visibility (i.e., the distance one can see) in the smoke in the duct. For purposes of comparison, visibility criteria used in performance-based evaluations of fire-protection systems are typically in the range of 5 – 25 m, depending on the type of building, the location in the building, and the familiarity of occupants with the means of egress.|
Table 1 . Detector response (analog output) relative to smoke optical density and visibility
|Smoke Optical Density (OD/m)||Detector Response (%)||Visibility*(m)|
|0.04 - 0.06||50||17 - 25|
|0.1 - 0.15||100||7 - 10|
*Visibility range for front-illuminated object.
Smoke consists of solid particles, liquid droplets, gases, and agglomerates of these three classes of matter. As smoke moves away from a fire source, it cools and changes characteristics, including smoke particle size, shape and colour. In addition, soot deposition on surrounding surfaces reduces the amount of smoke in the air stream. Until this research was conducted, little was known about how this so-called aging of smoke affects the response of commercial smoke detectors.
Researchers investigated smoke aging, measuring the number and size of smoke particles. They found that the number of small particles at the duct inlet compared to the number at a point 3 m downstream decreased by a factor of 10, while the number of large particles increased by a factor of two (smoke particles agglomerate and become larger with aging). This suggests that the aging of smoke occurs more rapidly than might be expected; in ductwork it occurs predominantly within the first few metres.5
Table 2. Effect of HVAC filter type on detector output
|Filter type||Dust-spot efficiency (%)||Decrease in detector output%|
|Photoelectric detector||Ionization detector|
|Group 1 (Glass fibres in a cardboard frame)||10 - 15||35||20|
|Group 2 (Extended area, pleated wet-laid cellulose)||30 - 35||55||40|
Previous research suggested that photoelectric detectors may be activated more quickly as a result of smoke aging than other types of detectors because their response is more dependent on particle diameter than on particle concentration. However, the full-scale experiments conducted in the NRC-IRC facility indicated that the signal output from all the duct smoke detectors was primarily dependent on the optical density of the smoke. The measurements were taken in the return air duct in the mechanical room, located a considerable distance from the fire compartment. By the time the smoke reached the measurement location, it had cooled to near ambient temperature and mixed with air and smoke from eight storeys in the facility. The results indicated that smoke aging did not affect smoke detection in the case of duct detectors.
HVAC filters remove some of the smoke as the contaminated air passes through them, with the amount removed varying according to the type of filter used. The performance of the filters is quantified through testing in accordance with ASHRAE standards6 , using the dust-spot efficiency measure to indicate the effectiveness of the filter in removing particulates from the airflow.
Measurements were conducted at the NRC-IRC facility in both the return air duct and in the supply air duct downstream of the filter. A comparison of the results from the two locations showed that the type of HVAC filter affected duct smoke detector performance (Table 2). The greater effect (decrease in detector output) on photoelectric smoke detectors is due to the fact that they are less sensitive to the small-diameter particles that pass through the filters.
The results suggest that the position, type, and efficiency of HVAC filters must be considered when evaluating the performance of duct smoke detectors. If used in return air ducts, the duct smoke detector is typically placed after all return air inlets, and before any filters, fresh air inlets, or fans (see Figure 1). Therefore, the filters would not have an effect on detector response. On the supply side, the duct smoke detector is placed after the fresh air inlet, the filter, the conditioning area, and the fan (see Figure 1). In this case, the filters would have an impact on the detection of fires (other than a fire in the filter itself), as they can reduce the smoke to a level at which it is not a concern. The project results indicate the duct detectors will respond if the smoke in the HVAC system reaches concentrations that are typically used as criteria for safe building evacuation.
Historically, standards have recommended that duct smoke detectors be located some distance (3 – 10 duct diameters) from bends, inlets and outlets in the duct to allow for uniform mixing of the smoke in the air stream. Practitioners have cautioned that long, uninterrupted straight runs of duct may cause stratification of smoke within the duct and thus negatively affect duct detector response (the concern being that stratification would concentrate the smoke at the top of the duct and could go undetected).
By taking measurements in a duct system near the fire source, researchers were able to demonstrate that at low velocities, buoyancy causes the smoke to concentrate in the upper part of the duct. But as the velocity increases, the distribution becomes more uniform, with the degree of stratification depending on the temperature of the smoke relative to that of the ambient air.5
Further experiments were conducted on a series of duct smoke detectors with both vertically and horizontally oriented sampling tubes on an extended length of duct in the NRC-IRC test facility's HVAC mechanical room, which was located remotely from the fire source. Thus the smoke had ample opportunity to cool, and there was no observable thermal stratification. The studies showed very similar performance for all of the detectors, regardless of location or orientation. Detectors located near bends and outlets had the lowest response (analog output). The results suggest that there is no justification for requiring duct smoke detectors to be located at large distances (3 – 10 duct diameters) from bends, inlets and outlets. It is recommended, however, that they be located at the mid-length of a straight run.
If the smoke temperature is above the ambient temperature, stratification in the duct can be expected until the smoke travels a sufficient distance to mix with the ambient air and loses energy to the surrounding environment. The distance at which stratification becomes a factor is dependent on the temperature of the smoke relative to that of the ambient air. This finding suggests that mounting smoke detectors in the upper part of horizontal ducts is the best approach. Duct detectors with sampling tubes may also be installed with a vertical orientation. In this case, the detectors should be mounted at the top of the duct to minimize the potential accumulation of dust in the detection system.
Efficacy of Sampling Tubes
There are two types of duct smoke detection. One uses spot-type smoke detectors installed in the ducts, which are the same as those installed on the ceiling. The other uses a similar detector contained in a housing attached to the exterior of the duct. Sampling tubes enter the duct to collect a representative sample of the air flowing through the duct.
Standard tests for duct smoke detectors include an evaluation of the sampling system at five velocities in the range of 1.52 m/s to 20.32 m/s. Because concerns had been raised regarding the performance of the sampling tubes being used, surveys of 65 commercial buildings in the Baltimore/Washington area were conducted. The surveys determined that the airflow velocities ranged from 2.06 m/s to 40.64 m/s, with only two of the HVAC systems exceeding the maximum air velocity stipulated in the standard test.5
Other experiments measured the response of duct smoke detectors as a function of system air velocities of 4.0 m/s to 19 m/s. Over this range, no significant variation in detector performance was observed. The sampling tubes were shown to be effective over the range of velocities typically found in HVAC systems.
The results of this research project indicate that
- the shutdown of the HVAC system will reduce smoke movement through the HVAC system;
- smoke dilution and smoke aging do not have an impact on the effectiveness of duct smoke detectors;
- the type of HVAC filter affects detector response, showing the greatest effect on photoelectric detectors, which are less sensitive to smaller diameter smoke particles;
- detectors in either the return or supply system will typically respond to smoke at concentrations at which safe occupant evacuation is still possible.
In addition, the results provide guidance on the location of duct smoke detectors and demonstrate the efficacy of sampling tubes for the range of velocities typically found in HVAC systems.
For further information on this research see References 7 and 8.
1. National Building Code of Canada, National Research Council, Ottawa, 2005.
2. National Board of Fire Underwriters, Smoke hazards of air-conditioning systems. NFPA Quarterly 33, 1939, p. 113-122.
3. CAN/ULC S524, Standard for the Installation of Fire Alarm Systems, Underwriters' Laboratories of Canada, Ottawa, 2001.
4. Mower, F.W., Milke, J.A. and Torero, J.L., A Comparison of Driving Forces for Smoke Movement in Buildings, Journal of Fire Protection Engineering, Volume 14, 2004, p. 237-264.
5. Wolin, S.D., Ryder, N.L., Leprince, F., Milke, J.A., Mowrer F.W. and Torero, J.L., Measurements of Smoke Characteristics in HVAC Ducts, Fire Technology, Volume 37, 2001, p. 363-395.
6. ANSI/ASHRAE 52.1, Gravimetric and Dust-Spot Procedures for Testing Air-Cleaning Devices Used in General Ventilation for Removing Particulate Matter, American Society of Heating, Refrigerating, and Air-Conditioning Engineers, Atlanta, GA, 1992.
7. NEMA, The Efficacy of Duct Smoke Detection, Fire Protection Engineering, Winter 2006.
8. NEMA, Duct Smoke Detection, Fire Protection Engineering, Spring 2006.
Dr. G.D. Lougheed is a principal research officer in the Fire Research program at the National Research Council Canada Institute for Research in Construction.
National Research Council of Canada