Fire detection in video sequences using a generic color model

In this paper, a rule-based generic color model for flame pixel classification is proposed. The proposed algorithm uses YCbCr color space to separate the luminance from the chrominance more effectively than color spaces such as RGB or rgb. The performance of the proposed algorithm is tested on two sets of images, one of which contains fire, the other containing fire-like regions. The proposed method achieves up to 99% fire detection rate. The results are compared with two other methods in the literature and the proposed method is shown to have both a higher detection rate and a lower false alarm rate. Furthermore the proposed color model can be used for real-time fire detection in color video sequences, and we also present results for segmentation of fire in video using only the color model proposed in this paper.     

A general formula for the prediction of vent flows

A new formula for fire-induced wall vent flow rate is developed based upon a theoretical derivation and mathematical fit to data. Previous research had developed a formula of mass flow rate for fire-induced doorway flows only. Here it is extended to include window flows. A theoretical model based on an ideal point source fire plume is used to guide the form of the empirical correlation. A thorough examination concerning the difference between the window and doorway flow modes is conducted. Both sill height and width of the windows pose key influence on the formula. The two vent configurations are merged into one equation. The results were compared to available flow data and shown to be within 15% accuracy for a wide range of fire conditions.     

Calculating combined buoyancy- and pressure-driven flow through a shallow,horizontal, circular vent: Application to a problem of steady burning in a ceiling-vented enclosure

A model was developed previously for calculating combined buoyancy- and pressure-driven (i.e. forced) flow through a shallow, circular, horizontal vent where the vent-connected spaces are filled with fluids of different density in an unstable configuration (density of the top fluid is larger than that of the bottom). In this paper the model is summarized and then applied to the problem of steady burning in a ceiling-vented enclosure where normal atmospheric conditions characterize the upper-space environment. Such fire scenarios are seen to involve a zero to relatively moderate cross-vent pressure difference and bidirectional exchange flow between the enclosure and the upper space. A solution to the problem leads to a general result that relates the rate of energy release of the fire to the area of the vent and the temperature and oxygen concentration of the upper portion of the enclosure environment. This result is seen to be consistent with previously published data from experiments involving ceiling-vented fire scenarios.     

Exploratory backdraft experiments

This study is a qualitative exploration of backdraft phenomena. Backdraft is defined as a rapid deflagration following the introduction of oxygen into a compartment filled with accumulated excess pyrolyzates. A scenario describing the physical and chemical fundamentals underlying backdraft phenomena is presented. A half-scale apparatus, designed to avoid dangerous overpressures, was used to obtain data from backdraft experiments. A gas burner supplied a 150 kW natural gas fire in a 1.2 m high, 1.2 m wide, 2.4 m long compartment with a small, 25 mm high, 0.3 m wide vent to ambient at floor level. Significant unburnt fuel accumulates in 180 seconds, when a hatch covering a 0.4 m high, 1.2 m wide vent, centered on a short wall, is opened, simulating a window breaking due to thermal stresses. The propagation across the compartment of the cold density-driven flow, which enters through the new opening, is called a gravity current. This gravity current carries a flammable mixed layer to a spark located near the burner on the opposite wall. The rapid deflagration that results upon ignition of the mixed layer is the backdraft. A compartment fire model is used to calculate conditions in the compartment before the vent opens. The hypothesized scenario appears to be confirmed by the deflagrations and exterior fire balls observed in these preliminary experiments.     

Comparison of Numerical and Experimental Results of Fire Induced Doorway Flows

The ability to calculate the changes to vent flows when a sprinkler activates can lead to improved predictions of fire environments outside of the room of origin in sprinklered occupancies, ultimately leading to an engineering design tool based on numerical simulations. Hence, for the current study, numerical calculations using NIST Fire Dynamics Simulator (FDS) are compared with real scale compartment experimental data for unsprinklered and sprinklered cases. Mass flow rate and temperature are typical parameters used to quantify the flow induced by a fire in a compartment. Hence, numerical results for doorway mass flow rate and temperature are compared with the experimental data for three fire sizes in order to validate the numerical model. Then, using current experimental data for sprinkler characteristics, numerical calculations for doorway mass flow rate and temperature are compared with the experimental data for the three fire sizes of the sprinklered case.     

Validation of ceiling jet flows in a large corridor with vents using the CFD code JASMINE

A field model code, JASMINE, has been adopted to calculate ceiling jet flows caused by plumes from unsteady fire sources in a large corridor. The idealized fast- and medium-fire energy release rates, obeying thet ² law for fire growth, were used. The effects of vents were studied by simulating three different configurations: a centrally located vent—that is, a vent directly above the fire source; an eccentrically located vent; and no vent. The results were compared with recent large-scale experiments conducted at the Swedish National Testing and Research Institute (SP) and with results from the computer program LAVENT. Some calculated sprinkler temperatures with three different RTI values are also presented.The present JASMINE simulations agree reasonably well with measured ceiling temperatures and velocities in large-scale tests. In general, the calculated ceiling jet temperatures are slightly overestimated and the velocities slightly underestimated. Also, the layer thicknesses calculated by JASMINE are somewhat thinner than those measured. The ceiling jet theory and present LAVENT results predict very thin layer thicknesses. However, these theories are only valid when wall effects are eliminated. From the present study it can be concluded that CFD models are generally more accurate when used to predict confined and unconfined flows.     

Thermal sources and fluid flow

It is shown that if air at a temperature T₀ enters a space at T₁, its effect on the space is that of a pure heat source of strength, Vs·(T₀−T₁), (Watts) where V is the volume flow in m³ s⁻¹ amd s is the volumetric specific heat in J m⁻³ K⁻¹. The flow is usually modelled as the conductance Vs, (W/K), but this is only correct if T₀ denotes ambient temperature; if T₀ represents the temperature of an enclosed space, the flow must be modelled as a source. Similarly, if a fluid flow is heated or cooled, it is to be modelled a temperature source, possibly together with a conductance.     

高大空间自然火灾温度场时空分布实验研究

开展高大空间大尺寸实体火灾实验,对高大空间温度场预测模型进行研究。通过对烟气温度曲线的最大温度、火灾进入衰退阶段的时间点以及曲线基本形状的构建,建立表达高大空间火灾全过程温度曲线的数学模型,并与已有的模型进行对比分析,发现三个模型均能够较好地预测高大空间自然火灾温度场变化及温度场的非均匀分布特征。提出的模型在火灾全盛阶段理论预测值与实验值吻合度较高,能够较好地反映火灾衰退阶段的温度场。     

A smoke detector activation algorithm for large eddy simulation fire modeling

This study chronicles the development and integration of a smoke detector activation algorithm (known as the SDAA) that describes the response time of a smoke detector into a large eddy simulation (LES) fire model [Roby RJ, Olenick SM, Zhang W, Carpenter DJ, Klassen MS, Torero JL. Smoke detector activation algorithm version 1 technical reference guide. NISTIR Report; 2006, in press]. Although the SDAA could be used with any CFD smoke movement model, the results here address specifically its application to the fire dynamics simulator (FDS). The fire model predicts the smoke concentration and velocity adjacent to the detector while an algorithm based on characteristic velocity-based lag times describes the transport of smoke into the sensing chamber of the smoke detector. The experimental data from a multi-room compartment fire were used for comparison and a series of benchmark studies provide a mechanism to establish the sensitivity of the model to the different input parameters. The SDAA was found to be very accurate in determining detector activation times for both high- and low-velocity smoke flows.     

Numerical model of the three-dimensional isothermal flow patterns and mass fluxes in a pitched-roof greenhouse

The three-dimensional isothermal flow patterns and mass fluxes in a full-scale, pitched-roof, single-span greenhouse were numerically resolved, and data from tests on a full scale were used to validate the code, the inlet boundary conditions and the greenhouse design grid method. For numerical solution of turbulent flow, a high-Reynolds-number k-ε model is suitable. Computational domain sizes were selected so as to fulfil the requirements of free-stream conditions whilst ensuring that grid geometrical characteristics satisfy the physical limitations of the standard k-ε model. A special feature of a case of a wind blowing parallel to a ridge (0°) is that the flow in the leeward half of the greenhouse comprises two vortexes with opposite senses of rotation, which bring in air mass through the vents and deliver it to the windward half. A spiral type of flow was found for winds blowing at 15-75° to the ridge direction: part of the air enters via the windward wall vent near the leeward gable-wall and emerges through the leeward roof vent near the windward gable-wall.Mass fluxes and flow patterns on wind direction, and on the opening angles of the windward and leeward vents. Thus, the ventilation rate induced by a wind directed perpendicularly to the greenhouse ridge is 4-4.9 times as great as that induced by a wind parallel to the ridge. A ventilation rate of a simulated greenhouse type was found to be significantly less responsive to a change in wind direction from 45° to 90° than to one from 0° to 45°. Present numerical results are in good agreement with those of other experiments and observations.     

A two-dimensional numerical investigation of the oscillatory flow behaviour in rectangular fire compartments with a single horizontal ceiling vent

This paper presents a comparison of fire field model predictions with experiment for the case of a fire within a compartment which is vented (buoyancydriven) to the outside by a single horizontal ceiling vent. Unlike previous work, the mathematical model does not employ a mixing ratio to represent vent temperatures but allows the model to predict vent temperatures a priori. The experiment suggests that the flow through the vent produces oscillatory behaviour in vent temperatures with puffs of smoke emerging from the fire compartment. This type of flow is also predicted by the fire field model. While the numerical predictions are in good qualitative agreement with observations, they overpredict the amplitudes of the temperature oscillations within the vent and also the compartment temperatures. The discrepancies are thought to be due to three-dimensional effects not accounted for in this model as well as using standard ‘practices’ normally used by the community with regards to discretization and turbulence models. Furthermore, it is important to note that the use of the turbulence model in a transient mode, as is used here, may have a significant effect on the results. The numerical results also suggest that a linear relationship exists between the frequency of vent temperature oscillation (n) and the heat release rate ( ) of the type , similar to that observed for compartments with two horizontal vents. This relationship is predicted to occur only for heat release rates below a critical value. Furthermore, the vent discharge coefficient is found to vary in an oscillatory fashion with a mean value of 0.58. Below the critical heat release rate the mean discharge coefficient is found to be insensitive to fire size.     

On the natural ventilation of two independently heated spaces connected by a low-level opening

We study the buoyancy-driven natural ventilation of two spaces which are connected to one another by a low-level opening, and each of which is connected to the exterior through a high-level vent. Each space is heated uniformly by an independent source, which provides buoyancy driving the ventilation. Using laboratory experiments, we show that these conditions lead to each space becoming well mixed at steady state. In this regime, a net flow from one space to the other is driven by the buoyancy created in the downstream space. Although it is possible in theory for the flow to develop in either direction, our new experiments and theoretical model show that, in reality, if the vents of the two spaces are at the same height, then the actual flow regime will depend primarily on the relative strength of the heat loads. If the two heat loads are sufficiently different, only the flow from the weakly heated space to the strongly heated one is stable. If the two heat loads are comparable, both modes are stable, leading to multiple flow regimes. The problem is generalised to show that, if the heights of the vents are equal, then the flow regime will depend on the relative height of the vents, as well as the relative strength of the heat loads. There is a range of combinations of vent heights and heat loads that still allow multiple flow regimes. We identify the limits of each regime and outline principles for control.     

A non-dimensional criterion and its proof for transient flow caused by fire in ventilation network

Application of 1D Transient Flow Model to a network analysis finds that the fluids in two branches (which are connected in parallel and contain no drive source) may no longer flow in a same direction while the network flow state changes from being steady into being unsteady due to a fire in the ventilation network. This phenomenon can be referred to as non-parallel flows in parallel branches. A non-dimensional criterion ξ is deduced theoretically. It is proved that for any two branches which are connected in parallel and contain no drive source, Non-parallel flows can be expected in the branches only if ξ is not equal to unity. If ξ is equal to unity, the non-dimensional flow velocities in the two branches will always be equal to each other. If ξ is not equal to unity, a prediction can be made on which one of the non-dimensional velocities in the two parallel branches will change more rapidly.     

The fire environment in a multi-room building— comparison of predicted and experimental results

This paper outlines results from a research project which is being used to investigate realistic fire environments in a prototype multi-room building. A comprehensive set of experimental data was obtained from a recently constructed three-storey Experimental Building-Fire Facility. The facility is used for a variety of fire investigation purposes, including fire growth and spread, smoke movement, and the effects of stair pressurisation and extinguishment. For the current investigation, a propane burner was located in the centre of a burn room to simulate a fire under both steady-state and transient-state conditions. The burn room was connected to other rooms. A comprehensive set of temperature, radiation and flow velocity measurements was obtained.

The numerical results obtained from a computational fluid dynamics (CFD) model were found to agree well with the experimental results. The CFD model results were also found to agree well with zone model predictions. These results encourage use of the CFD model to research the phenomena of realistic fire growth and spread and smoke movement in prototype building layouts.     

A numerical study of atrium fires using deterministic models

The smoke filling process for the three types of atrium space containing a fire source are simulated using the two types of deterministic fire model; zone model and field model. The zone model used in this simulation is CFAST (Version 3.1) developed at the Building and Fire Research Laboratories, NIST in the USA. The field model is a self-developed CFD model based on full consideration of the compressibility and k–ε modeling for the turbulence. This article is focused on finding out the smoke movement and temperature distribution in atrium spaces. A computational procedure for predicting velocity and temperature distribution in fire-induced flow is based on the solution of three-dimensional Navier–Stokes conservation equations for mass, momentum, energy, species etc. using a finite volume method and non-staggered grid system. Since air is entrained from the bottom of the plume, total mass flow in the plume continuously increases. Also, the ceiling jet continuously decreases in temperature, smoke concentration and velocity; and increase in thickness with increasing radius. The fire models, i.e. zone models and field models, predicted similar results for the smoke layer temperature and the smoke layer interface heights. This is important in fire safety, and it can be considered that the required safe egress time in three types of atrium used, in this paper is about 5 min.     

A study of non-flashover and flashover fires in a full-scale multi-room building

Realistic fire environments in a prototype multi-room apartment in a multi-storey building are studied. The fires are designed as non-flashover and flashover types, using standard polyurethane mattresses as fuel. A comprehensive set of experimental data is presented. The measured results include flame spread velocity, mass release rate, gas temperature, radiation heat flux and gas analysis. A computational fluid dynamics (CFD) model, called a CESARE-CFD fire model, has been used to simulate these polyurethane slab fires. The CFD model is described by three-dimensional transport equations for mass, momentum and enthalpy. The turbulence flow was modelled using the k−ϵ model. A soot formation model and a flame spread model were incorporated into the CFD model. The flame spread velocity and the mass release rate of the polyurethane slab fires were predicted in this study. It was found that the CFD model provided reasonable predictions of the magnitude and trends for the experiments both in the non-flashover and flashover fire cases.     

Gas dispersion near a cubical model building. Part II. Concentration fluctuation measurements

A wind-tunnel study has been performed of concentration fluctuations in the near-wake region (1.0 ? x/H ? 5.0) of a cubical model building in a simulated, neutrally stratified shear layer. Contaminants were released at a central roof vent for buildings with 0° and 45° orientations, and at a downwind roof vent for a building with 0° orientation.The log-normal concentration probability model was found appropriate for measurements in the building wake, and concentration fluctuation intensity was found to be reduced by the presence of the model building in an obstructed flow. A simple algorithm, based on the relation of the peak-to-mean concentration ratio to the local intensity, suggests an upper limit to the peak-to-mean concentration ratio near the ground centerline.     

Scale Tests of Smoke Filling in Large Atria

The large Atriums of airports and railway stations facilitate the access to transport vehicles including shopping malls, cultural spaces, etc. For this reason, they are used by an elevated number of passengers and visitors. Numerous malls contain a large atrium too, as a principal access or as a food court, and they usually have high occupant loads. In case of fire, the smoke can affect human health seriously, and people may be unable to reach a safe place before being overcome by the conditions created by the fire. The traditional approach to fire protection by compartmentation is not applicable to these large volume spaces and the ability of sprinklers to suppress fire in spaces with high ceilings is limited. This work evaluated—using scale tests, fire computer modeling and analytical methods—a comparative analysis of the different results obtained for the smoke control in large atria when the smoke filling approach is applied. Smoke layer and plume temperatures have been registered during the scale test—based on the Froude Modeling—and they have been compared opposite to the FDS scale simulation and the FDS large scale simulation. Smoke layer descend has been studied and compared for the scale test, the computer simulations developed and the empirical equations used. The results demonstrated that the evacuation time calculation is conservative when the zone computer model CFAST, the field computer model FDS or the empirical equations are used, although it turns out to be difficult to define the interface height based on the temperatures registered during the scale tests. The zone computer models generate results faster than field computer models or smoke tests, so it would be necessary to develop better calculation algorithms to define the smoke layer interface.

Critical ventilation velocity for multi-source tunnel fires

Ventilation is an effective method for controlling smoke during a fire. The “critical ventilation velocity” u_cr is defined as the minimum velocity at which smoke is prevented from spreading under longitudinal ventilation flow in tunnel fire situations. All previous studies on this topic have simulated fire scenarios in which only one fire source exists. This study conducted small-scale experiments and numerical simulations to investigate u_cr for cases in which two tunnel fires occur simultaneously. The tunnel was 4 m long, 0.6 m wide and 0.6 m tall. Three cases of two variously separated fires were experimentally explored and six cases were examined numerically. Both the experimental and simulation results indicated that for two identical fires, u_cr declines with separation. When the two fire sources are separate completely, u_cr can be determined by considering only a single fire. When the larger fire is upstream of the smaller downstream fire, u_cr also decreases with the separation. When two such fires sources are completely separate, u_cr can be evaluated by considering only the larger fire. The concurrent ventilation flow and flow of downstream smoke from the larger fire are strong enough to suppress the smoke flow from the smaller fire. However, when the smaller fire is upstream of the larger fire, the decrease in u_cr becomes insignificant as distance increases and the flow at u_cr must overcome the flow from both fires.