The use of numerical fluid mechanics to model smoke flow in buildings where a fire develops is common. It allows to check the effectiveness of ventilation systems at the design stage. It also gives the opportunity to determine the conditions that will be on escape routes. Numerical analyzes of smoke flow in buildings are most often performed using Fire Dynamics Simulator (FDS).

The paper presents numerical analyzes performed for the atrium building. The purpose of the calculations was to build a numerical model that corresponds to the real object located in the laboratory in Murcia, Spain. The analyzes consisted of fitting a numerical model based on the temperature distribution at selected points of the atrium. The model mapped the geometry of the real building and assumed the same fire power. Calculations showed high temperature compliance throughout the atrium, except for the vicinity of the fire source itself.

Buildings with a large open space are popularly called atriums. Atria are part of modern architecture. The architects design them as separate objects or they form a part of a building complex. It is a big challenge for engineers to create appropriate conditions for people to stay in such facilities. This applies to maintaining the right temperature and air quality. However, the biggest challenge is to ensure the safety of people in the event of a fire. When a fire breaks out there is a high probability that the entire building will be filled with smoke. This will cause difficulties in the evacuation of people. People in smoke lose their orientation and their speed of movement definitely decreases [

To reduce or even eliminate the uncertainty of the results obtained, there are many works aimed at the validation and verification of the FDS program [

The paper presents numerical analyzes made using the FDS program. The aim of the analysis was to match the numerical model to the results obtained in the real atrium building. For the validation of the numerical model, the results of the temperature distribution in the atrium located in Murcia, Spain were used. The atrium is in the laboratory of Technological Metal Centre. Many studies were carried out in the atrium and all of them were widely described [

To build the appropriate numerical model of the atrium, research presented by Gutierrez-Montes was used [

The pyramidal roof of the atrium was designed using nine steps. Each step was 0.5 m high. The construction of the roof is shown in Fig.

The roof and walls were modelled as 6 mm thick steel with a density of 7800 kg/m^{3}, specific heat of 0.46 kJ/kg K and conductivity of 45 W/K m. The floor was modelled as concrete with density of 1860 kg/m^{3}, specific heat of 0.78 kJ/kg K and conductivity of 0.72 W/K m. In the model of the atrium, square (1×1 m) pool fire was designed. The fire source was located in the centre of the atrium floor. The burning fuel in combustion process was heptane. Three different cases of the fire test simulation have been conducted. Each of 2.3 MW heat release rate fire. In all three cases, the same ambient conditions of the weather were adopted. The ambient air temperature of 16°C, pressure of 997 mbar and humidity of 49%. The default division of solid angles has been used. Other parameters have been left as the default values. In the simulations, the ongoing 900 seconds fire was investigated.

The computational domain includes the atrium space, the roof, the walls and additional space around the model of the atrium. In this paper, three cases of fire simulation are shown. Each case consists of different mesh sizes and different positions of the atrium model in the computational domain. In each case, different dimensions of the mesh cells were used. To achieve optimal simulation accuracy, mesh cells that are approximately the same size in all three directions were used. Table

Case | Heat release rate | Burning time | Dimensions of the computational domain | Dimensions of the mesh cells | Number of cells in each direction |
---|---|---|---|---|---|

1 | 2.34 MW | 900 s | 25×25×25 m | 0.25×0.25×0.25 m | X,Y,Z: 100 |

2 | 22×22×24 m | 0.2×0.2×0.2 m | X,Y: 110 |
||

3 | 25×25×25 m | 5 m around plume region: 0.125×0.125×0.125 m |
Near the plume region: X,Y: 40 |

In case 1 and 2 regular meshes were used. In these two cases, the main differences are dimensions of the mesh cells and the location of the model in the computational domain. In the first case, mesh with main dimensions of 25×25×25 m was used, while in the second case the main dimension of the mesh was 22×22×24 m. In the first case, the mesh had a dimension of the cell 0.25 m in each direction and in the second case that dimension was 0.2 m. Fig.

In the second case, the size of the computational domain has been reduced which allowed locating the model more evenly. Wall A was situated 1m from the computational domain boundary and wall C 1.5m from the opposite boundary. The remaining walls were located 1.5 m and 1m from the boundaries. In this case, the dimension from the highest point of the roof to the upper boundary was 2 meters.

In case 3, the computational domain was divided into two meshes. In the region above the pool fire and 5 meters around this region, the mesh with a cell size of 0.125 m in each direction was used. The remaining part of the computational domain was composed of a mesh with a cell size of 0.25 m in each direction. This combination was used to verify if thicker mesh around the plume region influences the results. The computational domain had dimensions of 25×25×25 m. The model was situated evenly from the boundaries. Wall A and wall C were located 2.75 m from the boundaries, and the walls without vents respectively 2.5 m and 3 m (Fig.

The most important numerical parameter in FDS is the mesh cell size. To verify how well the flow field is resolved, the non-dimensional expression D*/δx was calculated. According to McGrattan, the quantity D*/δx represents the number of computational cells spanning the characteristic diameter of the fire [

where:

^{3},

_{p}

^{2}

Table

Case | Cell size, m δx | Heat Release Rate, kW | Characteristic fire diameter, m D* | D*/ δx |
---|---|---|---|---|

1 | 0.25 | 2600 | 1.35 | 5.40 |

2 | 0.20 | 6.75 | ||

3 | 0.125 | 10.80 | ||

0.25 | 5.40 |

To study the thermal fields induced by the fire, several sensors were installed in the model. The arrangement of sensors was the same as in the real atrium [

The layout of the thermocouples measuring air temperature near the walls is symmetrical. As can be seen, both near the wall A and near the wall C, the sensors were installed at the same heights and the same distances.

During the numerical analyzes, attempts were made to adjust the model so that the results obtained during the simulation coincide with those obtained in studies on the real model. The calculations used temperature as a parameter whose compliance was tested. Figs.

The FDS results in three key regions are reported: the exhaust smoke temperature, the plume temperature and the temperature of the air close to the wall. The results of the three cases were compared to the results of the real measurements conducted in Murcia.

The results of the air temperature analysis in the exhaust vent are shown in Fig.

The Fig.

The Fig.

As can be seen the temperature rises at the height of 15.15 m first due to the smoke accumulation under the roof. (Fig.

Results of the analysis at the height of 5.15 m indicate the worst correlation with measurements in the real atrium. FDS over predicts the temperatures near the wall (Fig.

The paper presents the results of three numerical simulations carried out in FDS. Simulations have been conducted to check whether FDS can correctly determine the temperature distribution in the atrium model. The calculation results have been compared with the results obtained from measurements in the real atrium. Three cases of the simulation were conducted to check the influence of the grid size on the results.

Calculations show good agreement with the results of the measurements in the atrium, except the plume region near to pool fire. The space above the source of fire is a place of very intense turbulent flows. The best compatibility of the results of measurements and numerical calculation for the temperature was obtained for a uniform mesh of 0.2 m. Creating a denser grid around the fire source did not improve the numerical results. Incompatibility of the results of measurements and numerical calculations appeared above the fire source regardless of the density of the mesh.