Experimental and CFD Modeling of Fire Development in a Small Compartment Effect of Ventilation - Term Paper Example

Experimental and CFD modeling of Fire Development in a small compartment – effect of ventilation. Introduction Fire is an extremely uncontrollable combustion process, which involves the laws of thermodynamics, kinetics, fluid dynamics, physical sciences and many more incidences. A set of Mathematical equations can be used to explain this process, notwithstanding its complex nature. Such mathematical models can be used to estimate the occurrence of dangerous situations, and to take corrective measures on time. Using of mathematical models to solve problems associated with fire has its origin from analytical models. The behavior of fire can be predicted using a variety of models during post-flashover period as well as during pre-flashover periods. Computer fluid dynamics (CDF) models is a commonly used. Validation of CFD is critical, if its codes are to earn credibility - this builds confidence in its predictability. Validation is particularly essential, when making codes for complex situations – which involves impossible experimental tests, or demanding in terms of cost, time or efforts. Use of CDF models for fire and smoke control and management has been described in many literatures. The complexity of this modeling has, however, raised concern and has not been appreciated by some engineering groups. Notably, the users who have some slight knowledge in CDF model theories can be able to apply it, due to its increased user-friendliness. To demonstrate this, Yau et al (2003) has explained how a volumetric heat source can be sued to obtain good results, rather than using combustion model. This is achieved by appropriate selection of the volume and the area of the fire. This guidance can be used in a variety of simulations associated with CFD models. Methods The use of CFD simulation to solve difficulties involves the stages below Calculation domain – The capacity is determined from which the combination is acquired. In some cases, the calculation domain is known while for others the calculation amount should be cautiously chosen for precise forecasts. Framework creation – A framework is built in a manner that guarantees that an absolute combination is acquired. Where the number of framework is high; the calculation time takes a lengthier time. A framework that results to precise and reasonable solutions ought to be created. Some framework attributes like aspect proportion may have an effect on the mixture and should be cautiously scrutinized in the initial stages. Borderline Settings – At the borderline of the calculation area, it is important to ascertain heat and fluid condition. In situations where there is unrestricted solution movement, borderline settings should be ascertained away from the movement to ensure the mixture is not altered. Fire Modeling - This is accomplished by allocating a volumetric heat source at the control capacities where burning takes place using a burning replica. Caution should be taken in specifying volumetric heat sources as this ensures accurate attributes of smoke are forecasted. To determine the effect of the above on forecasting, the CFD simulation FDS developed by NIST (McGrattan et al 2002)] is used to replicate several situations. The situations are chosen to permit scrutiny on the effect of replica parameters on the solution. The replica outcomes are then scrutinized and charts generated, which evidently show this effect. Where feasible, the outcomes of the simulation are evaluated with present relations. Grid resolution In a CFD simulation, the use of small mesh elements usually increases the accuracy of the results. Spatial gradients are approximated using more points and a greater proportion of the flows are modeled for large-eddy simulations without having to settle for a turbulence model.  The major setback of higher resolution meshes is their increased cost of computation. While selecting a mesh resolution, one which produces results with the desired accuracy should be considered while taking into account limitations on availability of time and computational resources. Mesh refinement is the most important in the combustion volumes of plume boundaries and layer interfaces when it comes to CFD simulations of fire cases.  To accurately model the production of heat and fire products, combustion models will need enough fine mesh resolutions. Moreover, since areas of relatively high gradients in the flow are represented by plume boundaries and layer interfaces, small mesh elements are required to correctly predict entrainment and mixing rates.   In addition, grid resolution affects the accuracy of the mixture fraction model used in FDS as usage of coarse mesh elements in the combustion volume can lead to under predicted heat release rate and flame heights (McGrattan & Forney 2004). Also, FDS makes some changes to the mixture fraction model to cover for these effects. On the contrary, grid resolution has a small effect on the accuracy of volumetric heat source modeling, but at the same time volumetric heat sources are usually poor at predicting heights. n this case, the presented simulations are variations of the ‘plume3’ model supplied with FDS Version 3 (McGrattan & Forney 2004). A 0.2m square burner located in the center of a flow domain, 1.6m wide by 1.6m deep by 3.6m high (Figure 1). For the smallest mesh element, the size is 0.025m, three distinctive mesh elements are used for a 24 kW fire, and three heat release rates (Table 1).  The vertical mass flow rates and plume centerline temperature profiles constitute the major comparison. Table1:Plume3 simulation parameters The point where an equivalent point makes a similar plume as a fire is the virtual origin. This constraint is commonly applied in assessment of plume dynamics of fire. It is estimated as follows. z0 = -1.02 D + 0.083 Q2/5 (1) The heat release rate in this case is represented by Q and the diameter in meters is D (SFPE 1995). The area of the burner is 0.04m2 in all models, and the burner perimeter is matched exactly with the grid. The corresponding thickness of the burner, which is equal in all sides, is calculated as follows. D = (4A/ π) 1/2 = 0.2257 m (2) The hypothetical elevation of the flame, L, is estimated as follows, assuming atmospheric conditions do not change (SFPE 1995). L = -1.02D + 0.235Q2/5 (3) The comparison of the predictions of simulations in regard to the plume centerline hotness is as shown in figure 2. Here, simulation A passes through C, against the elevation (z), which can be illustrated by the following relationship. T0 = 9.1 (T∞ / g cp2 ρ∞2)1/3 Q c 2/3 (z-z0)-5/3 + T∞ (4) In this correlation T∞, cp and ρ∞, are the temperature (K), heat capacity (kJ/kg-K), and the density (kg/m3) of the ambience air, and the fire heat release rate (kW) part of convection is Qc. The rate of flow for the 24 kw simulations can be compared with the hypothetical correlation, based on the following equations, which stands for the weak plume for axisymmetric. m = 0.153 (g ρ∞2/cp T∞) 1/3 Qc 1/3 (z-z0)5/3 (5) Investigational correlations are shown by equations 4 and 5. As such, these tools are considered very exact when comparing arithmetical modeling results. These correlations are also developed for accurate measurement of square fire sources, besides axisymmetric fires. Comparison of simulation’s plume centreline temperature profiles is as shown in figure 2. The rate of flow of mass for simulations C, D and E is as shown in figure 3. For a steady part, the results reveal that, as the fire size increases, the simulated profiles decreases. This is in accordance with Ma’s and Quintere rules that provides that the correctness of plume modeling relies on the fire size non-dimension, Q*, which can be illustrated as follows: Q* = Q / (ρ∞cpT∞D2 (gD)1/2) (6) Figure 3 shows the comparison of plume 3 rate of mass flow, with different sizes of fire. The three profiles that have been simulated here reveal that there is a departure near the top domain. The size Q* which is non dimensional can be used for comparison of grid resolutions for various heat release rates, as explained in length by Quintiere and Ma (Ma & Quintiere 2003). Framework Analysis Usage of reduced netted components in a CFD replica enhances the precision of the outcome. Spatial gradients are estimated by means of additional locations for large-eddy replicas, a larger amount of the moving currents are replicated right away without turning to a turmoil replica. The disadvantage of increased analysis netting is its amplified calculation expenditure. A framework analysis should be selected, which produces outcomes with the anticipated precision, but also considers the limits on available time and calculation means. For CFD replicas of fire situations, netting enhancement is vital in burning capacities, smoke borderlines and heat stratum borders. Burning simulations need netting analysis to precisely simulate creation of thermal and fire outcomes. Smoke borderlines and stratum borders signify zones of reasonably elevated gradients in the solution movement, and need tiny netted components to precisely forecast entrainment and combining proportions. Framework analysis also influences the preciseness of the solution portion simulation applied in FDS. The application of uneven net components in burning capacity has resulted to lower forecasted thermal discharge proportions and flame altitude (McGrattan & Forney 2004). FDS creates certain alterations to the combination portion simulation to counteract these results. The disparity is that framework analysis has less impact on the preciseness of volumetric heat source simulation, which is usually weak in foretelling flame altitudes. Replicas depicted in this matter are variants of the ‘.Plume 3’ simulation provided with FDS version 3 (McGrattan & Forney 2002). A 0.2 m four sided gas jet is placed in the middle of a movement area 1.6m in width, 1.6m in depth and 3.6m in height (Figure1). Three dissimilar netted components magnitudes are employed for a 24kW fire and three thermal discharge proportions are employed for the tinniest netted component magnitude of 0.025m (Table 1). The major assessment parameters are the vertical mass movement speed and smoke counter line hotness profiles. Foyer Plume Resolution Modeling fire plume dynamics is an essential prerequisite for fire safe engineering designs. Simple plumes can be examined using standard correlations. CFD techniques permit examination of plumes which lack standard shapes or else plumes that interrelate with multifaceted geometries. Mesh resolution is the major modeling concern that is necessary to perfectly calculate plume dynamics. Plume analyses demand an approximate entrainment rate which affects the plume range, mass current rate, temperature and the deepness of the hot air created in an enclosing volume. Perfect calculation of entrainment plume ends requires modeling eddies down to a specific bare minimum magnitude. A great eddy simulation model as in FDS, translates to a minimum mesh component dimension since smaller eddy than it will be projected to a certain extent than modeled. The presented simulations are of plain fire in foyer. The foyer is 20 m2 and 30m in area and height respectively, without a wall and an upper limit. The exhaust aeration is not specified. On the foyer floor is a 4MW fire, spread over 9 m2 centered. Two pieces of meshes for simulation D in tern allow high resolution of the burning volume. Mesh 1 has been defined as a 6m Cube around the fire with the sides open and the top; Mesh 2 being defined to be the remainder of the foyer. The said Mesh 2 is modeled to 24m high by 20m square. The outstanding simulations use a single mesh above the entire foyer. The mesh essentials are isometric and all simulation solved to a condition. Major parameters for distinction are the plume centerline temperature profiles and the vertical mass course. Figure 5 compares the flow rate profiles for the four Simulations where the theoretical profile is based on equation four. The outcome of simulation D are fairly close to the theoretical mass flaw velocity and the nearest of the four theoretical plume centerline temperature outline. It is expected as simulation D, which has the highest resolution of the burning volume. Great resolution of the burning volume improves the perfectness of the outcome of the high elevations in comparison to the ½ m single mesh simulation E, with noticeable discontinuities in Profiles of simulation D at the boundary between the two meshes (6 m). The abrupt variance in mesh size may cause modeling errors. Simulation G is one of the three simulations that use a single mesh and it yields profiles contiguous to theoretical because of the high mesh resolution that it holds. The inconsistency between outcome from simulation F (0.25 m) and G (0.2 m) indicate that 0.2m is neither a grid-insensitive resolution for this representation. Modeling at higher mesh resolutions would be mandatory. High resolution of the burning volume is not accurate in prediction of the bulk plume course at advanced elevations. To produce result of computational economy, the grid sensitivity studies ought to be performed to certify the model’s correctness. An estimate account suggests that the mesh resolutions of the order of 10-1 m ought to be used for plume models. In the recent past, Ma and Quitntiere (2003) have published a mesh sensitivity study functional to plume dynamics modeling. Aspect Ratio Solution times can significantly decrease with the use of a non-isometric mesh, which happens to reduce the number of elements in a flow domain. Non-isometric elements are particularly appropriate where there is a strong directed flow in the direction of the element transformation. With high advection which tends to decrease spatial gradients in the direction of flow - the need for fine mesh elements along one axis is reduced. The accuracy of the results will be altered, since highly non-isometric mesh elements may introduce numerical errors into simulation. Table 2: Parameters for ‘tunnel’ simulations Simulation Aspect ratio Mesh 2 Number of Element Size Elements(m) Total Solution Time (min) Mesh 2 Solution Time (min) Tunnel A 1.1 40 0000.5 x 0.5 x 0.5 785 59.31 Tunnel C 1.5 8 0000.5 x 2.5 x 0.5 703 7.93 Tunnel D 1.1 4 0000.5 x 5.0 x 0.5 683 3.98 The presented simulations in the table 2 above are 60metres long, 10metres square tunnel open at one end with a 5 MW fire covering 9m square next to the closed end. The burner is resolved with a 10m cubic mesh (Mesh 1) using a comparatively fine resolution of 0.25m to accurately model the combustion process and plume dynamics. The remainder of the tunnel (Mesh 2) is resolved at 0.5m down the tunnel’s width and height, and four dissimilar aspect ratios down the tunnel’s length: 1:1 (0.5m), 1:5 (2.5m) and 1:10 (5.0m). The table also lists the solution times for the three simulations and it is clearly seen that, as the number of elements in Mesh 2 are reduced, the total solution time decreases. Though not presented in table 3 above, the mass flow rate profiles along the tunnel were compared for all three simulations indicating how well entrainment into the hot layer is modeled as it flows along the tunnel. Simulation A serves as a basis for comparison for the other two profiles, despite it being grid-sensitive. What is more, simulations C and D both demonstrate a large increase in the mass flow rate along the mid portion of the tunnel, thus indicating larger entrainment rates into the hot layer. In addition, due to propagation of the open boundary condition into the flow domain, both profiles show a high amount of unbalanced variation near the end of the tunnel, and this increases the aspect ratio, moving the variation farther into the flow domain. Figure 6 compares the temperature profiles at the tunnel exit for all three simulations. The results of these simulations describes the careful application of non-isometric mesh elements with selected comparison against results using smaller aspect ratios, even though they allow for savings in solution time. This data also supports the close that, improved aspect ratios artificially increase entrainment rates into the hot layer, resulting in a thicker layer with the rise of temperature being low. Consequently, this will affect the specifications of the tunnel heat detectors and more significantly, the estimate of the smoke exhaust system capacity will increase. Doorway Jet Plumes Fire engineering professionals are always the lot concerned with compartment fires and they believe that accurate modeling of compartment fire dynamics is vital in predicting such parameters as secondary ignition, time to flashover and smoke and heat production from the source. Through a radiation heat transfer and interaction with the flow of combustion air into a compartment, the doorway jet plume can influence conditions within a compartment. In order to reduce computation time, it is desirable to minimize the size of the flow domain. Also, concern is the radiation that will be lost from the flow domain and interaction with the combustion air layer, which will be poorly predicted if the plume is not modeled for a sufficient distance beyond the door. As a result, modeling of the doorway jet plume is inevitable to achieve accurate results. Table 3: Parameters for ‘door-plume’ simulations Simulation Distance from door (m) Mesh element Size (m) Doorplume A 0.0 0.0625 Doorplume B 0.5 0.0625 Doorplume C 1.0 0.0625 Doorplume D 2.0 0.0625 Presented in table 3 above are simulations of simple compartment fires with varying distances outside the room being included in the flow domain. In all simulations is a mesh size of 0.0625 and they are being run for five minutes to achieve a stable-state. Located at the floor level centered against the wall opposite the doorway, is a 1m2, 500 kw burner and in the flow domain are four different distances outside the room; 0.0, 0.5, 1.0 1nd 2.0m (Table 3). The compartment is 2.5m width, 3.0m length and 2.5m height with a 1.5m width and 2.0m height doorway centered in one of the shorter walls (Figure 7). Assessment of profiles B, C and D show little change in the temperature profiles as the modeled plume length is increased from 0.5 to 2.0 beyond the door indicating that even if only a short length of the doorway jet plume is modeled, the grid-sensitive results may be obtained after all. Since temperature profiles are to be compared throughout the compartment, figure 8 compares the door centerline temperature profile for all simulations. The outcome from simulation A is noticeably different from the rest, with a higher flame temperature and increased hot layer elevations in the room and at the doorway opening, but this has partly been caused by poor modeling of the complex flow dynamics at the door opening. In addition, the arrangement of an open flow domain boundary at the doorway that satisfies the boundary conditions changes the flow pattern across the door. And lastly, the loss for radiation from the doorway jet plume back into the compartment may possibly affect the compartment temperature profiles. Temperature profiles were compared throughout the compartment. Figure 8 compares the door centreline temperature profiles for all four simulations. Results from simulation A is markedly different from the rest, with a higher flame temperature and increased hot layer elevations in the room and at the doorway opening. This is partly due to poor modeling of the complex flow dynamics at the doorway opening. The specification of an open flow domain boundary at the doorway changes the flow pattern across the door in order to satisfy the boundary conditions. The loss of radiation from the doorway jet plume back into the compartment may also be affecting the compartment temperature profiles. Comparison of profiles B, C and D show little change in the temperature profiles as the modeled plume length is increased from 0.5 to 2.0 m beyond the door. This indicates that grid-insensitive results may be obtained even if only a short length of the doorway jet plume is modeled. Interface Height Fire safety engineering designs requires approximate of layer crossing point height all the way through the confined space. The zone models are usually designed to compute these layer crossing points directly counting the average temperatures in both the hot and cold layers. On the other hand, the CFD models compute temperatures all through the course domain. Physically, a reasonable method is thus necessary to compute the three values as indicated in the CFD data. Prior models of FDS tolerated temperature data to be gathered at separate points which were either processed in the same manner as experimental data. This approach demanded considerable data processing. The FDS model 4 uses a numerical combination based on maintenance of mass and power to process perpendicular temperature profile into the mono-interface height. The FDS also computes both the cold and the hot layer standard temperatures. Time is usually the varying quantities but the three values are usually available. There is CFD software that makes the use of the k-E turbulence model and it performs poorly when forecasting the layer heights. This is attributed to the variances among the large eddy simulations along with the k-e turbulence model. On the other hand the LES models ably and directly model the proportion of the course eddies (Hadjisophocleous, Lougheed, & Cao, 1999). Modeling Soot FDS soot models is transported like gaseous fumes. The soot yielded from resources is particularized as constants. The simple management avoids performance of an additional expensive smoke model. Note, there are issues related to smoke modeling and the same are not addressed by the FDS and the include; Particle soot agglomeration: this is where the soot particle tends to stick on each other over duration of time. This result in particles of larger size though less. There are also errors in the determining the particular size of the soot and this may crash on occupant ability to visualize and foretell efficiency of photoelectric and the ionization smoke sensors. The issue of soot particles being deposited on solid surfaces is also one that raises concern. Particle size: the size of the soot particle size varies with the nature of fuel and the burning conditions. The soot particle size is a fundamental factor in determining smoke obscuration, thus the visibility of the occupant and the sensor efficiency are affected. Separation Flow: The FDS models the soot as a continuous segment, as opposed to articles of mass. This dealing does not foretell any partition of the soot from the course as it changes bearing. This may be serious for simulations of sensors. The FDS has a semi-solution in the sense that it allows separate soot particles to be modeled as Langrangain. Colour: The color of soot particles primarily depends on the fuel type for instance hydrocarbons give much darker soot as compared to Cellulosic soot. The FDS fails to allow soot particle color modeling. For instance, when the smoke obscuration is determined as purely light absorption by the particles of the soot, their color fails to have effect. The reality is that light diffusion for light colored smokes severely decreases the effect of proper visibility due to poorer light consistency. The reaction of the smoke sensors is based on backscattering technology, which gets difficult to evaluate the FDS results. Layer Boundry Modeling Perfect modeling boundary layers are essential for thermo fluid simulations. It is noteworthy that a Fluid’s peripheral velocity is at zero while at a solid surface and the same time increases to the freestream velocity though what is referred to as boundary layer. The boundary layer model, when affected by the fire simulations is affected as the transfer of the heat up from the hot burning gases to surfaces that are solid. Usually, this influences the secondary explosion times, pyrolysis rates, high temperature losses from chambers and the improvement of upper limit and layer walls as they cool. The boundary layer modeling can affect the entrainment into the ceiling jets. The significance of boundary layers differs with the nature of the fire incident. The compartment fire is quite distinct, the hotter layers can move faster thus facilitating an accurate modeling of the boundary layers fairly insignificant. In the lower speed course, like the ceiling jets originating from the Plumes intensifying over large distance from the ground, boundary layers cause an immense impact on the course. Scenarios characterized with low speed courses as the sensor activation modeling may be predisposed to errors in layer boundary. The FDS models boundary layer improvement is by an application of a peripheral velocity boundary at fluid-solid interfaces. Where there are fumes, most boundary layer thicknesses are usually in the range of 10 to 3m. Such a thin resolution for construction-scale if difficult due to the high calculation cost. The FDS sets for the fluid velocity at a solid interface to compensate for the lack of mesh resolution. As a fraction of the value of the velocity in the subsequent mesh component away commencing the wall. Heat and mass transfer at the fluid-solid are handled majorly from an empirical correlations. The FDS permits a straight arithmetical simulation (DNS) to be conducted where all course features are computed straight devoid of resorting to turbulence or eddy model. At a solid interface the fluid velocity is usually zero. The modeling of boundary layer comes with benefits like; more precise calculation of mass and heat transfer. Regardless of the requirement of mesh resolution is perfect, thus limiting the dependency of DNS to small scale simulations for majority of modelers. When scrutinizing the grid sensitivity study of the boundary layer, the boundary layer may be physically small; however, the boundary layer has the potential to impact the rest of the flow. Conclusions This paper seeks to illustrate the use of the CFD models in application associated to fire modeling and smoke progress in enclosures and tunnels. The literature demonstrates and illustrates the impact on the explanation of parameters used in defining and resolving the dilemma in these models. Critical parameters examined include: resolution of the grid, entrainment plume rates, models of combustion, smoke models, fraction radiation, , height calculation interface, and boundary layer modeling. These illustrations have informed the following conclusion: That the precision of the plume distinctiveness for a constant mesh element decreases with an increase in fire size. The end result from the ‘foyer’ simulations show that higher resolution model of burning volume cannot necessitate a perfect foretell of the bulk plume course at high elevations. Therefore, a rough guideline, the mesh resolutions in the order of 10-1m ought to be used in plume models. The non-isometric mesh components have an impact on the result and ought to be applied cautiously. Characteristic ratios build up the impact on the solution turn out to be considerable. Courses though windows doors ought to be perfect foretell with the totaling of the computational domain of small length. In instances where non-orthogonal grid is necessary, it was established that velocity compensation in FDS makes a difference in the flow dynamic around non-orthogonal geometries. The radiative part affects significantly fire improvement and time for the flashover for a partition should be selected keenly to represent the fuel burning. In instances where the fuel is single object consisting of multiple fuels, the package ought to be modeled as a block constituting fuel properties computed as an average mass weighed. It should be noted that the approach that is preferred if that of direct modeling of fuel package geometry and constituents if sufficient calculating resources are accessible. References Hadjisophocleous, GV, Lougheed, GD, Cao, S 1999, Numerical study of the effectiveness of atrium smoke exhaust systems, ASHRAE Transactions, Vol.105, no.1, pp. 699-715. Ma, TG, Quintiere, JG 2003, Numerical simulation of axi-symmetric fire plumes: accuracy and limitations. Fire Safety Journal, Vol. 38, no. 5, pp. 467-492. McGrattan, K, Forney, G 2004, Fire Dynamics Simulator (Version 4) User’s Guide, National Institute of Standards and Technology, July 2004. McGrattan, K, Forney, G, et al. 2002, Fire Dynamics Simulator (Version 3) User’s Guide. National Institute of Standards and Technology, November 2002. SFPE 1995, Handbook of fire protection engineering, 2nd Ed. National Fire Protection Association. Yau, R, Cheng, V, Yin, R 2003, Treatment of fire source in CFD models in performance based fire design, International Journal on Engineering Performance- Based Fire Codes, Vol. 5, no. 3, pp. 54-68. Read More