The prescriptive approach to designing an Aspirating Smoke Detection, or ASD, system is very well established. In fact, many projects simply position sampling holes in the same place as point detectors and, by using a reliable approved detector, the designer is assured that the minimum performance requirements will be satisfied. It is also widely recognised that high sensitivity ASD systems are capable of providing much earlier warning than conventional technologies – particularly where smoke is diffused and enters more than one sampling hole. Indeed, the early warning capability has been frequently tested and demonstrated using the early warning performance tests, such as the 2m PVC overheated wire test defined in the FIA (formally BFPSA) Code of Practice for ASD systems, and in NFPA 76 in the United States.
As a result of the Regulatory Reform (Fire Safety) Order there is increasing emphasis in the UK on risk based assessment. This is steering the UK industry away from a rule based “deemed to satisfy” approach towards a performance based design and fire engineering approach. While this is a good thing, there is one small problem. The current performance based design approach for ASD systems involves assessing the system with a test such as the hot wire test after installation. While this is effective for separating the best system designs from the mediocre or poor, it is totally reliant on the empirical, the experimental and on the previous experience of the designer. So how do we overcome this problem?
The field of fire engineering has been steadily advancing throughout the second half of the last century. Far from being limited to academic research, many of the tools developed are now widely available to, and used by, the mainstream fire engineering community.
Some of the most important tools for fire modelling use Computational Fluid Dynamics (CFD) to assess system performance against the design criteria. One of the most popular of these tools is the Fire Dynamics Simulator (FDS), developed by the US National Institute of Standards and Testing (NIST). With this and other such tools it is now possible, even in reasonably complex environments, to predict the amount of heat produced by a particular fire, how gas moves within the area, and the distribution of smoke and heat at any stage during the fire event. Of course, these models are dependant on the accuracy of the assumptions and data entered but with the increasing power of computers, the complexity of models possible is proving sufficient to yield some really useful predictions.
Traditionally, CFD modelling was used to determine the alarm response time and smoke obscuration levels for fires that were already well developed; the results for large fires being reasonably reliable. More recently, however, we are beginning to see some success in incipient (smouldering) fire modelling, and the modelling of situations where the effects of a fire would be dominated by background conditions – for example, a draft caused by an open door at the end of a corridor. Despite the fact that CFD smoke concentration prediction accuracy continues to improve, due to differences in smoke entry characteristics and detection technology algorithms, the predicted alarm response times are heavily dependent on the type of detector being modelled. Fortunately for us as a company, the smoke entry characteristics of our Vesda ASD devices are extremely easy to predict; even more so when the system is supported by Aspire2 – Xtralis’s pipe network modelling software tool.
To be specific, an ASD system uses an aspirator (or fan) to draw air into the system. As such, pipe modelling software, like ASPIRE2, can be used to predict when smoke entering each and every sampling hole will reach the detector. This type of deterministic model is not available with point detectors, for example, where the smoke entry time is much more dependant on the inter-relation between the design (for example, the insect screen) and external flow conditions, such as ceiling jet velocities and normal air flows. There is much evidence that the smoke entry characteristics of such ‘passive’ point detectors are unpredictable. ASD ‘active’ sampling benefits from greater predictability in a range of applications – evidenced by the reliability of the Aspire2 modelling.
So, by linking the outputs of the CFD fire models with the smoke entry and alarm thresholds of the ASD, it is possible to obtain reasonably accurate predictions of the alarm time. This can then be incorporated into the fire engineered performance based design solution for the particular application environment.
Fire engineered solution
Xtralis has investigated the use of CFD modelling for a real smoke detection system design on several occasions now. However, one of the first was undertaken at a time when there were no reports on using FDS to directly evaluate the performance of a smoke detection system, so the design was also validated by real smoke tests. This provided an excellent opportunity to compare our predictions with real data.
The project centred around a 60m high atrium divided into protection zones. The atrium was modelled because of the detection difficulties presented by large open spaces: long distances for smoke to travel to reach detectors resulting in smoke dilution; ventilation which has the same effect; the possibility of smoke stratification; and the potential for no smoke plume formation.
From our research, a preliminary set of validation CFD models were generated and used to verify the suitability of the simulation parameters that were used in the actual atrium models. Specifically, a number of fuel types, fire sizes and room geometries were simulated and compared with the results of validation tests. These preliminary models indicated that differences between the simulated and real test ASD alarm response times would be no more than 20% and, therefore, be within industry accepted limits.
In the atrium simulations, four fire locations were investigated, all being in the middle of a set of four sampling points, and so representing the worst case scenario of the furthest distance that smoke would need to travel.
Real hot smoke tests were conducted to validate the design, using the same parameters as for the CFD modelling – fuel, fire size, fire location, and so on. The detectors in the zones closest to the fire reached ‘Fire 1’ within 60 seconds, which compared well to the alarm response times in the simulations. Using this data, we were able to fine tune the system design to comply with local requirements, such as maximum permissible response time.
Suppression release
Another area in which we use CFD alarm response time predictions to assist with the engineering of real fire solutions is with automatic extinguishing systems. There is already a large number of applications where high sensitivity aspirating systems are being used for early warning smoke detection, to avoid unnecessary actuation of suppression systems. This is not surprising since the cost of the ASD system is actually a relatively small investment when compared to the cost of an unwanted alarm and suppression dump – the cost of the suppressant and the clean up process can run into thousands.
While there has been a lot of interest in fire suppression systems, little emphasis has been placed on when it is most appropriate to release the fire suppressing agent. Decisions regarding the detected smoke level at which systems should be triggered for the maximum impact of the suppression agent are not widely understood, and it is a common assumption that normal point detectors provide triggering at the best time – not too soon where the agent is wasted on a small fire, and not too late when the fire is too large to control. This is, perhaps, why the more general term is fire suppression as opposed to fire extinguishing.
What is of interest is that alarm signals from aspirating systems are increasingly being integrated into the suppression release logic. Not only is the high sensitivity of aspirating devices being used to provide early warning of a fire, but the normal sensitivity alarm levels that are considered to be equivalent to traditional point detectors are being used for suppression actuation. The ‘Fire 2’ alarm threshold, available on Vesda laser-based detectors is ideal for this purpose. In fact the latest release of
BS 7473-1:2006 no longer recommends against the use of ASD systems within a co-incident release strategy, while guidelines issued by VdS in Germany specifically deal with using such systems for extinguishing release, making it clear that as long as two separate detectors are involved in the decision to release, they may both be ASD devices.
Interestingly, neither of these documents make any specific reference to the sensitivity requirements; it is assumed that any aspirating system used in suppression release will be set up with sensitivities similar to a point detector. Clearly, using a system with a Class C alarm threshold according to EN 54-20 is one of the simplest ways to demonstrate equivalence to point detectors. In this case, if smoke of a concentration sufficient to trigger an
EN 54-7 optical point detector enters one sampling hole, then the system will signal an alarm to the extinguishing control panel. In many situations this is sufficient, particularly where the number of sampling holes is small (typically less than 10). Where there are many sampling holes within the same area, it must be considered that the suppression system could theoretically be triggered if smoke of a much smaller concentration enters all sampling holes simultaneously. If, for example, there are 10 holes then smoke of one tenth of the concentration than required to trigger an EN 54-7 optical detector might signal an alarm if it enters all 10 holes! For this reason, it is often the case that the second co-incident alarm input signal does not come from another ASD device. However, this theoretical situation is very unlikely, and what is really needed is a better understanding of how smoke will move within the protected area, to better predict the response of detection systems.
Using the fire system designer’s knowledge of the area to be protected and the designer’s experience of smoke propagation, it may be reasonable to estimate which groups of sampling holes would become exposed to smoke, and then desensitise the alarm threshold accordingly. In fact, Aspire2 has a sampling hole ‘group function’ to assist with the calculations necessary to determine the appropriate thresholds when making these design decisions.
More recently, we have been using CFD modelling to more accurately predict the concentrations of smoke entering each sampling hole. While this can be done on a project by project basis, it is often too expensive to perform the calculations needed to obtain a unique solution for an individual project. So, in conjunction with some leading organisations in the US, we have developed a generic calculation tool. This is known as the ASD Suppression Alarm Threshold (ASAT) calculator. This was recently approved by FM Global and is derived from a series of CFD simulations and verification experiments in typical suppression environments. By simply entering the area and height of the protected area – plus some information on the number of air changes per hour and intended detector spacing – the calculator provides an estimate for the ‘Fire 2’ threshold needed to achieve equivalence to point detectors.
Summary
ASD technology is already well established for both prescriptive designs, with proven equivalence to point detectors using detectors with appropriate approvals, and performance based designs, when tested using appropriate performance tests like those defined in the FIA Code of Practice. More recently, with the increasing reliability of CFD fire modelling and the predictable smoke entry characteristics of ASD systems, it is becoming possible to accurately predict the response time of systems in real applications. Two good examples of how this has been applied to practical applications of ASD are the protection of large open spaces with high ceilings, and integration with co-incident automatic suppression release systems. It is anticipated that this emerging capability will lead to further improvements in the quality of the design process for fire detection systems.
Peter Massingberd-Mundy is technology and expert practices manager at Xtralis (formerly Vision Fire and Security) and sits on CEN TC72 working group 16, which developed the new EN 54-20 standard for aspirating smoke detection.
