Effectiveness of sprinklers for fire suppression 24 Aug 2017

Only two countries currently have prescribed water flow design guidelines for deluge systems in their motorway tunnels: Japan, which requires a 6mm/min flow rate⁽¹⁾ and Australia, with a 10mm/min flow requirement⁽²⁾. The substantial difference between these requirements indicates the uncertainty of the underlying fire science.

Fig 1. Schematic of the test setup

To quantify the performance of a deluge system and provide information for the optimization of these otherwise suboptimal systems, full-scale tunnel fire experiments were conducted to establish the absolute minimum required water flow required to supress an in-tunnel car fire.

The experiments involved placing a burning car under a single BETE deluge sprinkler nozzle type N9W20.4 positioned 5.5m from the road deck (Fig 1). Thermocouple trees were placed at the centre and at both sides of the car to estimate flame height, assess fire severity at the target surface of the tunnel lining and to determine the risk of fire spread before and after sprinkler activation. Additionally, thermocouples were placed inside the car, under the roof, between the front seats, in the middle of the rear foot compartment and in the middle of the backseat to further gauge the heat release. Video recordings enabled understanding of the different stages of burning behaviour and infrared video cameras estimated flame heights when visibility was lost. Based on this footage, flame heights were determined and the heat release rate was estimated.

The water flow for two forms of deluge fire suppression, gradual and instantaneous, was simulated and controlled by regulating the system’s water pressure (Fig 2), a relationship expressed in the equation

To determine the distribution of the water flow at ground level, a series of standard ‘pan tests’3 were performed where pans were placed along the radius away from the sprinkler nozzle. After a certain discharge time, the volume of water in each box was measured, divided by the surface area of the box (356mm²) and divided by the time to determine the water distribution. This process was repeated for a range of operational water pressures. To assess the radial symmetry of the water distribution, additional pan tests were performed for two pressures in 45 degree increments to cover the entire surface.

To estimate the heat release rate (HRR) of the flame height from the burning car, the mean flame height (corresponding with a probability of 0.5) was measured by collecting the flame height of 100 infrared frames (4sec at 25FPS) and developing a cumulative normal distribution. This is necessary because the intermittent part of the flame fluctuates several times per second and an instantaneous analysis will not resemble the HRR of the fire.

Fig 2. Pressure/water flow relationship

Heskestad’s flame height correlation⁴ was used to derive heat release rates from the mean flame heights. This relationship is given in the equation Q ̇=((L+1.02D)/0.235)^(5⁄2) where D is the diameter of the fire (m) and (Q ) ̇is the heat release rate (kW). The flame length L (m) is derived from the infrared camera recordings and the fire diameter D is assumed to be the width of the car.

The main fuel load in a car consists of the seats, interior lining, plastics in the bodywork, the tires and the fuel in the tank (which was not present for this experiment). Additional fuel from cabling and other combustibles in the engine compartment are added to the fuel load. The car doors were removed prior to the test and placed inside to maintain the fuel load.

A bag of woodchips drenched in diesel was used as the source of ignition and placed in the middle of the foot compartment of the backseat. While not necessarily the most likely scenario, it represents the worst possible case.

For assessing burning behaviour of the car, the focus lies on thermocouples T1-T5 (Fig 1). If the temperature in the immediate vicinity of the car is reduced, it is assumed that a drop of at least this magnitude occurs further away from the burning vehicle, for example at the tunnel walls. An instantaneous drop in temperature is defined as a drop to a constant temperature below 100° during the 100sec of sprinkler activation. A gradual drop is a drop that occurs over a longer period of time. If one temperature drops slightly, but another one rises, the temperature in the immediate vicinity of the car is assumed to rise.

• 

  Setup of water distribution test

• 

  Fig 3. Water distribution away from the deluge nozzle

• 

  Fig 4. Representative water flow rates for the car area

For the variation of the water dispersion along the radius away from the sprinkler nozzle, a trend from a single peak for the highest system pressure towards a double water flow peak for lower pressures can be deduced from the variation of the water dispersion along the radius away from the sprinkler nozzle (Fig 3).

To find the representative water flow for a certain system pressure, it was assumed that a car fire is roughly equivalent to a pool fire with a diameter of 2m, meaning that the water flow values (Fig 4) must be integrated from the origin to the 1m radius mark. Since the integration of the water flows is carried out for a radius equal to unity, the resulting numerical value is the representative water flow or the average water flow (Fig 5).

Five tests were performed with deluge activations at different water flows (Figs 6, 7, 8, 9 and 10).

A drop in temperature during the 100 seconds of sprinkler activation is divided into three main categories: a gradual drop in temperature that occurs steadily when the system is active, an instantaneous drop in temperature when the temperature reaches a constant temperature below 100°C within 100 seconds (subdivided to include the effect of the steam pushing down the flame, which delays the instantaneous drop next to the car) and a rise in temperature after deluge activation (Fig 11).

Data points corresponding to these categories correspond roughly to as many regions, but the number of data points is insufficient to categorize the plot in three distinct regions. The transition from one sprinkler effect to another is therefore marked with shaded regions. The most conservative approach is to assume that the worst scenario would occur in this region i.e. temperature rises in the horizontal shaded bar and gradually rises in the diagonal region).

Fig 11. Graph of the effect of sprinklers to suppress vehicle fires in tunnels

Above 6.6mm/min, the water flow should drop temperatures around a burning car. Whether this drop is instantaneous or gradual depends on the water flow that is applied. Essentially, there are two ways of achieving an instantaneous drop in temperature: to either increase the water flow or keep the fire small in size by minimizing sprinkler activation time.

To design a deluge system for high traffic density tunnels, eliminating the potential for fire spread relies on dropping the temperatures in the immediate vicinity of the car. If the expected tunnel traffic density is not that high and the lanes are not particularly narrow, fire spread to adjacent cars might be less relevant and design could focus on keeping temperatures at the tunnel interface at the tunnel walls below a certain threshold.

The tests indicate that there are two ways of achieving an instantaneous drop in fire temperature: by either applying a sufficiently high water flow or by limiting the size of the fire. The former leads to a more conservative, expensive system whereas the latter comes down to minimizing sprinkler activation time. This emphasizes the importance of rapid deluge activation.

The boundaries between the three regimes depend on the flow of water that is applied. A water flow of 6.6mm/min managed to drop the temperatures around a burning car. Below this the fire is assumed to be out of control. These results are only valid for the exact configuration as tested, with no ventilation and for the given fire sizes.

Whether an instantaneous or gradual drop in temperature is required depends on the distance to an adjacent car. This directly relates to the expected traffic density and the traffic lane width.

In closing these tests, it is a given that further research is needed to investigate the effects of the potential rise in temperature next to the the burning car following deluge activation and that additional analysis on scalability is required to investigate applicability on heavy-goods vehicles and on the influence of droplet size and application of longitudinal ventilation.