Reconstruction of the Grenfell Tower fire – Part 4: Contribution to the understanding of fire propagation and behaviour during horizontal fire spread

The tragic events at Grenfell Tower in 2017, involving a combustible façade system, have raised concerns regarding the fire risk that these systems pose. In this series of articles, so far published, fire development inside the initial apartment has been investigated using an appropriate computational fluid dynamics (CFD) model. Several scenarios including different fire sources and ventilation conditions were addressed. Fire propagation through the window to the external façade and to higher apartments was modelled. This model was validated by comparing the numerical results with the visual observations reported in the Grenfell Inquiry. A CFD model of the complete east face of the Grenfell Tower was then created. This paper details CFD modelling of the complete Grenfell Tower façade during the late horizontal phase of fire spread. As the physics of lateral flame spread is different from that for upward flame spread, it is important to assess the validity of the model, thus far developed, for this configuration. Fire propagation over the whole façade is modelled and compared with observations from the real disaster. This provides a better understanding of its fire behaviour and of the contribution of architectural details and their impact on fire spread.


| INTRODUCTION
The Grenfell Tower is a 24-storey high-rise building located in In previous studies, 6,7 a multi-scale numerical approach was used to assess the fire performance of the refurbished façade system used on the Grenfell Tower. This was underpinned by intermediate scale ISO 13785-1 8,9 and large-scale BS8414-1 7,10 reaction-to-fire tests and used the computational fluid dynamics (CFD) code fire dynamics simulator (FDS) version 6. [11][12][13][14][15] The output of these simulations showed that the ACM cladding was the main product driving the overall fire behaviour of façade construction in this specific system. In particular, systems that featured ACM cladding made with a polyethylene core (ACM-PE) showed extensive fire propagation regardless of the insulant used. 7,16 Fire spread over facades is one of the fastest modes of fire spread in buildings. The excess of air, the verticality of the façade, especially when no deflectors are provided, and the inside/outside pressure difference are some of the main factors that facilitate fire spread, and it spread from compartment to compartment. [17][18][19][20][21] The development of the fire, inside the kitchen of the apartment of origin at Grenfell, and its behaviour at the kitchen window were investigated in Reference 22. Fire propagation over the façade and from the façade to apartments through windows was also explored. The overall heat release rate (HRR) for typical apartment rooms and window failure criteria were estimated roughly, based on assumed apartments contents prior to the fire. A complementary thermomechanical analysis of window failure was performed and reported in Reference 23.
The full height of the Grenfell façade was modelled numerically to determine its fire behaviour. 16  This publication reflects the full-scale fires on the east, north, south and west faces of the Grenfell Tower, using the numerical models previously validated for the façade system, the initial apartment fires and window failure. Fire spread over the external façade and to higher apartments was validated by comparison with video and photographic records of the real fire. 5 This provides a better understanding of fire behaviour and the role of architectural details in fire spread. Additionally, the numerical model allows further investigation into the effect of changing the insulant material used in the cladding system.

| The Grenfell tower
Architectural details of the façade system and the Tower are indicated in references. [2][3][4]22 The Tower perimeter included a series of 14 columns: five columns on the north and south faces of the building leading to four bays, and four columns for the east and west faces of the building, leading to three bays as shown in Figure 1. Hence, respectively, the north and south faces, and the east and west faces were identical. From levels 4 to 23, all floors had a similar layout of six flats (four two-bedroom flats and two one-bedroom flats) and a lobby, as indicated in Figure 1. These flats are called "X1" to "X6" in this paper.

| Numerical setup
The three-dimensional CFD model reported in Reference 16 was adapted to simulate horizontal fire spread over the east and later the north, south and west faces of the Tower following the initial vertical propagation.
In the numerical model of the Tower, it is assumed that there is no fire propagation pathway between floors inside the Tower via ducts, HVAC systems or holes in apartment ceilings or walls. Fire propagation from one apartment to another, horizontally or vertically, occurs only via the façade and depends on window failure. Wind was not considered in the model because the analysis described in Reference 5 showed that the horizontal fire propagation rate over the Tower was equal in both clockwise and counter-clockwise directions, at the level of the architectural crown and at different heights above the 19th floor, and was, thus, independent of the wind. Therefore, this work considers horizontal fire spread on the upper parts of each of the four faces of the Tower independently.
Because the east and west faces, and the north and south faces, respectively, are identical ( Figure 1), two different numerical models were built to simulate fire propagation over the faces of the Tower The emissivity is taken as 0.87. The window units incorporated metalfaced XPS cored sandwich panels. The thermal properties of XPS are taken from References 25-27. Namely, a density of 20 kg/m 3 is used in this study. The heat capacity and thermal conductivity are, respectively, 1.13 J/g/K and 0.03 W/m/K and the emissivity is taken as 1. The ignition temperature of XPS protected by a metallic sheet can be estimated as around 500 C. 25 XPS has a heat of combustion of 40 MJ/kg. 26,27 The CO yield is 0.06 g/g 26,27 and the soot yield is 0.2 g/g. 26 The asymptotic mass loss rate is 0.032 kg/m 2 /s. 26 29 The CO yield is 0.063 g/g, 26 the soot yield is 0.176 g/ g 26 and the HCl yield is 0.27 g/g. 26,30 The asymptotic mass loss rate is 0.016 kg/m 2 /s. 28 The thermal properties for mineral wool are extracted from Efectis database and from product datasheet. It has a density of 360 kg/m 3 and a specific heat of 1.0 J/g/K. The emissivity of the cavity barrier is taken as 1.
The actual architectural details related to window frames, aluminium rails and cavity barriers have been used. The window failure criteria assessed in references 22,23 are used for each apartment opening throughout the façade, that is, partial window breakage when a surface temperature of the frame reaches 550 C 31 ( Figure 3).
The HRRs of the initial fire, as well as the apartment fire contributions, further detailed in Reference [22], are assumed for each floor of the Tower. The HRR of the individual furniture implemented in the apartment is indicated in Reference 22 and an overview is addressed in Figure 3. Individual pre-defined heat release curves are used for kitchens, living rooms and bedrooms. The same curves are used for all flats on all floors, in the absence of more specific detail. The total surfaces of these rooms are conserved in the numerical model. For the two bedroom apartments, "X2," "X3" and "X5," the HRR per unit area was the same as the "X6" flats because of their identical layout. For the one-bedroom flats, "X1" and "X4," the HRR per unit area of the "X6" flats was kept. Thus, the HRR of each "X1" and "X4" room varies according to its surface area. It was assumed that these compartment

| Horizontal propagation over the whole tower
The fire spread over the four faces of the Tower was modelled from 01:29 to 04:30 AM The progress of the numerical simulation, every 30 minutes, is shown in Figure 6.    Table 1 addresses the synthesis of the horizontal fire propagation at crown level in terms of spread rate at each location. Although the average spread rate for both clockwise and anticlockwise directions is similar, the local values at specific flats or at columns are quite different. Indeed, the average spread rate at corner columns is 60% higher for the clockwise direction than for anticlockwise. However, the local effect of the columns included in the façade tends to decrease the average spread rate for the clockwise direction but to increase it for the anticlockwise one. However, the average horizontal spread rates for each face of the Tower, including the architectural details, is close to 0.3 m/ min (Table 2) and to the values taken from observations and detailed in Reference 5.  Table 3.
Numerically, the angle of the V-shape is higher for the east face and then decreases for north, south and west faces. Following observations, a similar angle is found for the east and north faces, while a lower value is evaluated for the south and west faces. The angle related to the establishment of the V-shape seems to decrease with the order of the faces involved in the fire (east then north, south and west) and thus follows the horizontal spread around the Tower. This can be related to the fire being more intense during its first phase (vertical spread and horizontal spread along the east face) mainly driven by the combustion of the cladding.

| Heat release rate
The total HRR estimated numerically during the fire spread over the four faces of the Tower is shown in Figure 17. However, no data are neither used nor available to compare the HRR. It is essentially a result from the simulation.

| INVESTIGATION OF THE EFFECT OF A CHANGE IN THE INSULATION MATERIAL
The impact of using a non-combustible insulation material was investi-  is consistent with that evaluated in Reference 5. However, although the average horizontal rates were comparable, the columns included along the faces seem to decrease slightly the local horizontal spread at crown level in the clockwise direction and to increase the spread rate in the anticlockwise direction, meaning that the fire propagation rate over the façade between columns was influenced by the direction. During the horizontal spread across the east and north faces, the simulation showed that the fire was clearly propagating in a V-shape and an average angle close to 0.7 /min was calculated for both faces.
However, the angle related to the establishment of the V-shape seems to decrease with the order of the faces involved in the fire (east then north, south and west) and thus follows the horizontal spread around the Tower. This can be related to the fire being more intense during its first phase (vertical spread and horizontal spread along the F I G U R E 1 9 Horizontal fire propagation at crown level -Clockwise spread direction from observations 5  Even if the numerical model addressed in this paper correlates well with observations during the Grenfell fire, several modelling assumptions were needed as detailed in previous studies in this series.
The same considerations can be addressed for the present paper, in