Reconstruction of Grenfell Tower fire. Part 2: A numerical investigation of the fire propagation and behaviour from the initial apartment to the façade

The dramatic event of the Grenfell Tower in 2017 reminds the importance of addressing fire issues as a whole and clearly highlighted one of the major roles played by the façade as fire propagation vector. To understand and analyse this disaster, numerical simulation allows particular phenomena to be evaluated more easily. The numerical model addressed for the fire behaviour of the façade system was developed using a multiscale approach and validated at different scales. In this paper, the fire behaviour of the façade and of its window frames is addressed. A computational fluid dynamics (CFD) model is used to investigate the fire spread from the initial apartment to the overall façade with different scenarios for the fire source and ventilation. Fire propagation through windows to the façade and to upper apartments is addressed. General curves representing the re‐entry of flames into upper apartment are extracted from simulations. The numerical results are validated by comparison with observations from videos and pictures of the real fire. Numerical results show that whatever the initial fire location and ventilation conditions, even if the fire source is of hundreds kilowatts, it is enough to ignite the adjacent element early and to the appearance of external flames shortly after.


Summary
The dramatic event of the Grenfell Tower in 2017 reminds the importance of addressing fire issues as a whole and clearly highlighted one of the major roles played by the façade as fire propagation vector. To understand and analyse this disaster, numerical simulation allows particular phenomena to be evaluated more easily. The numerical model addressed for the fire behaviour of the façade system was developed using a multiscale approach and validated at different scales. In this paper, the fire behaviour of the façade and of its window frames is addressed.  4 were performed with the code Fire Dynamics Simulator (FDS) 5 to reproduce three of these fire tests. Thermal characteristics of the system components were integrated into the model. The justification of the numerical model used for thermal degradation analysis was addressed. The first aim of that study was to validate the numerical model by comparing its output with the aforementioned experimental results, including details in terms of flows and thermal conditions at each location in the tested system.
The second aim was to allow a better understanding of how the fire propagates on the overall system and on the insulation and cladding, of which the cladding system comprises. The fire behaviour of each component of the overall system was thus modelled numerically.
After the Grenfell disaster, the UK government commissioned seven large-scale BS 8414-1 fire tests. 6 The objective was to determine the combinations of insulation and ACM cladding that could safely be used. In Dréan et al, 7 the influence of scale on the fire behaviour of façade systems featuring ACM-based cladding was investigated numerically using the validated numerical model described in. 4 Simulations were performed in order to reproduce three of the UK Government's BS8414 fire tests, detailed in. [8][9][10] Then, the numerical model was modified to use coarser numerical cells that are more commonly encountered in large-scale simulations and engineering studies.
Using such a coarse grid was necessary because of the difficulties and time taken in modelling larger scales, such as a full-scale façade on a high-rise building. However, a numerical hypothesis must be fixed in order to apply the model developed using an accurate fine grid to a coarser grid. The main objective was to reproduce the thermal gradients in gas and solid phases achieved with the initial model. Several studies have shown the feasibility and usefulness of numerical simulation for high-rise building fires [11][12][13] and for post incident analysis. [14][15][16][17] Numerical simulations were used, after the World Trade Centre disaster in 2001, to provide deeper analysis of the fire propagation inside the structure. 14 One of the most important features for assessing the fire behaviour of façade systems is to properly represent fire spread from the fire compartment of origin to the façade. External flames venting through an opening, such as a window aperture, expose the façade system to heat and can lead to ignition of façade components and fire spread to the façade. 18 This, in turn, creates a significant risk of fire spread to adjacent floors or buildings.
The performance of façades, when exposed to flames venting through a compartment opening, has been extensively studied, experimentally and numerically. [19][20][21][22][23][24][25][26][27][28][29][30] The Grenfell Tower is a 24-storey high-rise building. The initial fire, before spreading outside, was a localized fire close to the wall and window, in the south-east corner of the kitchen in flat 16, located on the fourth floor of the east façade of the Grenfell Tower. The CFD model was used to evaluate the spread of the initial kitchen fire, to the building's façade through/around the kitchen window opening and the propagation of the fire over the façade. Re-entry of the fire into subsequent apartments through window openings is also assessed, and general fire evolutions representing the re-entry of the flames inside the apartment are extracted from the simulation for the kitchen, the living room, and the bedroom. The model was validated by comparing its output with video and photographic records of the real fire. Additional thermomechanical analysis was performed on the windows used in the tower refurbishment. Both open and closed windows were analysed, to determine whether they would fail by falling inwards or outwards. Thermal loads from apartment and façade fire simulations were considered as boundary conditions. Predicted time to first failure of the frame was around 4 to 5 minutes. This paper presents the hypothesis selected and the results for each step of the study performed.
Regarding the uncertainties related to the initial fire source that has led to the Grenfell disaster, the main objective of the present study is to provide a better understanding of the possible fire scenarios that lead to the development of the fire along the façade. Due to the complexity of the investigation and the lack of information, numerical models are useful to provide elements or scenarios that can be discarded. More complex models can be proposed if more information are available, and several scenarios can be assessed numerically. However, in this work, a preliminary analysis of the probable fire development in the initial apartment is addressed, until the ignition of the façade system.

| DESCRIPTION OF THE FAÇADE SYSTEM AND OF THE INITIAL APARTMENT
The Grenfell Tower is a 24-storey high-rise building, refurbished in the

| NUMERICAL SET-UP
In this study, a 3D CFD model is used to evaluate fire spread from the initial apartment (flat 16), located on the fourth floor of the east façade of the Grenfell Tower, to the façade through the kitchen window. The initial fire within the kitchen of flat 16 before spreading outside was a localized fire close to the kitchen's wall and window.
Interaction between the wall and the fire must be taken into account for a meaningful assessment of fire spread. Thermal action by the fire on the initial apartment's walls, particularly on the wall corresponding to the façade, is not homogeneous. A strong gradient exists, including the window frame. A zone model will give an average temperature that may underestimate local temperatures, in particular, in this case, at the window end of the kitchen. 3D modelling assesses local effects and these can be compared with video and photographic observations.    José Torero's expert report. 33 This study assumes that the first ignited item is located in the south-east corner of the kitchen, close to the window, as proposed in Professor Luke Bisby's expert report. 31 Regardless of fire source, the development of the fire will be controlled by ventilation and hence the flow rate of air through the kitchen's window and doors. A fridge/freezer and a mini-fridge were reported as being present in the kitchen of flat 16. Thus, numerical modelling has been performed considering each fridge as the fire source. The smouldering of both fridges is represented by a slow increase in HRR over 5 minutes.
Thus, during the first 5 minutes of the fire, the global HRR is less than the 300 kW discussed in Professor José Torero's expert report. 33 Three fire scenarios are investigated based on Professor Luke Bisby's expert report 31 and Professor José Torero's expert report 33 (Table 1).
In scenario 1a, the fire source is the mini-fridge existing at the left of the kitchen's window. The kitchen's door is kept closed in this scenario, assuming the confinement of the fire by the owner and by the firefighters. In scenario 1b, the fire source is the same, but the effects of the door opening is investigated, representative of the firefighters entering several times in the kitchen. The scenario 2 investigates the change of the fire source, since the fridge/freezer is also designated as a potential source.

| Scenario 1a: ignition of the mini-fridge with restricted ventilation
The development of the fire from the mini-fridge in the kitchen of flat 16 under restricted ventilation conditions is investigated. The sliding partition between the living room and the kitchen, and the doorway

| Scenario 1b: ignition of the mini-fridge with free ventilation
For scenario 1b, the same apartment configuration and furniture as in    The fire behaviour inside the kitchen and the living room ( Figure 10) shows that between 9 and 11 minutes, the fridge/freezer ignites. After

| Scenario 2: ignition of the fridge/freezer with free ventilation
As it is unclear what the first ignited item in the kitchen was, except that it was located in the corner of the kitchen close to the column and the window corner, a second fire scenario was investigated.
In this case, the large fridge/freezer ignites first. The evolution of HRR for this scenario is presented in Figure 12 and compared with the HRR from scenarios 1a and 1b. A similar evolution of HRR is observed for both locations of the fridge. The main difference between HRR behaviours in scenarios 1b and 2 is that higher values of HRR are reached locally in scenario 2, due to the proximity of the fridge/freezer to the other kitchen contents. The window failure begins earlier, 12 minutes after ignition.
In Professor Luke Bisby's expert report, 31 it is noticed that the first observation of fire having spread to the cladding is at 01:09:00 AM, corresponding to 15 minutes after the first emergency call, and 19 to 20 minutes after the fire ignition. This is coherent with the numerical observations indicated in Figure 13, with a first cladding ignition around 19 minutes. Even if the fire starts being extinguished inside the apartment at 01:20:00 AM, the propagation at the façade has begun.

| Preliminary synthesis
For the three fire scenarios investigated for the initial fire development in the kitchen, whatever the fire location (fridge or mini-fridge, scenario 1b vs 2) and ventilation level (scenario 1a vs 1b), the same conclusions can be drawn.
• Even if the fire source is less than 300 kW, it is strong enough to ignite the adjacent PVC window surround early.
• The fire propagation quickly leads to the ignition of the fan unit mounting panel and to the appearance of external flames quickly.
• When surroundings elements are ignited, the window frame fails in a few minutes.
• Even if the fire starts being extinguished inside the apartment at 01:20:48 AM, façade fire propagation has begun.
• Flashover in the kitchen is not reached in any scenario.
The synthesis of the times for first window failure and cladding ignition is addressed in Table 4.    The numerically evaluated HRR for the living room is shown in Figure 19. Its contribution is negligible. The first peak is due to the dining table ignition, and the second peak is due to the partial ignition of the sofa. Flashover is not reached.  However, it is assumed that these bedrooms are identically furnished and that the fire will develop in a similar way for each room. Thus, the fire HRR achieved is considered to be applicable to the two bedrooms in each apartment.

| Numerical evaluation of window failure
To verify the failure delay achieved during the CFD analysis, the ther-

| CONCLUSIONS
This study aimed to model and to understand the fire development and propagation inside the apartment of origin of the Grenfell Tower fire and the behaviour of the kitchen window of that apartment. The The fire behaviour at window frames and its spread through the façade was investigated. The synthesis of the times for first window failure and cladding ignition was addressed for the three scenarios.
Fire evolution for scenarios 1a and 1b were comparable; thus, the window failure seemed to be the main ventilation vector and the doors aperture appeared to have a reduced influence. More accurate representation of the kitchen furniture, with detailed modelling of individual apparels, could also be a further investigation.
Modification in the thermal properties of the materials and in the fuel definition will be evaluated in a later research. This work must be considered as a preliminary analysis of the probable fire development in the initial apartment, until the ignition of the façade system.