Reconstruction of Grenfell Tower fire. Part 1: Lessons from observations and determination of work hypotheses

The Grenfell Tower fire occurred on 14 June 2017, killing 72 people. The pattern and speed of vertical and horizontal fire spread characterize this catastrophic event. Plentiful video and photographic data of the fire spread available has been carefully verified and concatenated into a database. The verified data have been superimposed on a projection of the Grenfell Tower in order to track the development of the fire. The surface that is unburnt, burning, or extinguished, as well as the presence of internal fire at any given location, is thus recorded for the duration of the fire. An analysis of the results showed that the initial vertical propagation can be divided into three phases. After the façade ignited at the fourth floor, vertical propagation over time is linear, with a vertical fire spread rate of around 3.5 m/min until the fire reached the sixth floor. Then fire propagation decelerated. Finally, fire spread accelerated with a power four dependence. The maximum vertical fire spread rate was around 8 m/min as the fire reached the crown at the top of the building. Horizontal spread proved to be greatest at the level of the crown (0.293 ± 0.005 m/min). There is a linear relationship between speed of horizontal fire spread and height. These correlations and observations yield important conclusions, and eight different hypotheses capable of explaining the global behaviour of the fire are suggested.


| Floors design and elevation
The model tower's floors are all identical, from fourth to 23rd. Each floor comprises six apartments. Two of the apartments are only exposed to one side of the tower (the east and west face, respectively), whereas the remaining four apartments are corner apartments meaning that they are exposed to fire on two faces of the tower.

| Observations
The available evidence tends to suppose that the fire started in the kitchen, close to the window, of apartment 16, on the fourth floor of Grenfell Tower. The fire burned within apartment 16 for more than 10 minutes before breaking out onto the external façade of the tower.
The fire was notified to the emergency services at 0:54:29 AM.
Fire is visible at the kitchen window of apartment 16, from approx- The photographic and video evidence between the beginnings of the façade fire (approximately 1:08:06 AM) and the fire reaching the crown (approximately 1:29:00 AM) shows that the fire was contained between two vertical columns on the east façade. Table 1 summarizes these data.

| Data processing
The observations set out were plotted onto a graph (  Hypothesis 2. The initial flame spread is driven by convection, which is driven by linear modes.

| Data processing
The observations set out in Table 2 have been plotted in Figure Table 3. Lane confirms there appears to be a correlation between the levels to which external water was applied and the lack of external damage on these levels (paragraph 17.5.7, page 32). Without this external firefighting, it is probable that fire will propagate downward more, eventually to ground level.

Data processed Observations
Between 2:34:00 AM and 2:48:00 AM, the fire is clearly propagating in a V-shape down the tower. The fire also reaches the south face of the tower. There is evidence of downward propagation along vertical columns of the east face of the tower that remain intact, ie, the fire has been successfully constrained between two vertical columns.
Between 3:08:00 AM and 3:11:00 AM, the fire reaches the west face of the tower. Propagation continues at the same rate, especially at crown. These observations can be summarized as below.
• Post fire, many windows frames were still in place or partially in place with only their upper parts were melted.
• The outer and inner parts of the window frame and casement frame were split, meaning that the thermal break had been destroyed, leading to a loss of integrity in the window.

Data processed Observations
At 4:09:00 AM, the clockwise and anticlockwise fire paths merge on the west face, close to south-west corner.
After the fire extinguished, some parts of each façade are unburnt. The boundaries of these unburnt parts may demonstrate the limit of external firefighting efforts. • A large part of the upper preframes were melted. The lower preframes were protected by the insulant and so were often in better condition than the upper profiles.
• A lot of windows had fallen into the apartments-if windows were tilted open at the time of the fire, they may have had the moment to fall in an inwards direction.
• Some windows were still fixed to the lower pre-frame.
It can be difficult to determine whether the state of the windows is due to early failure or prolonged exposure to fire. Hypothesis 3: During this period, the fire enters apartments 26 and 36 on floors 5 and 6, respectively. The deceleration in the rate of spread is linked to the time it took for the fire to move into these apartments and to reach flashover. The additional energy from these internal fires burst back out to the external façade of the tower and restarted the vertical spread but at a higher rate.
Observation 4: The rate of flame spread accelerates to the power four between 1:21:00 AM and 1:29:00 AM.
Hypothesis 4: The fire spread is driven by radiation from the fire plume from 1:21:00 AM.
Observation 5: Throughout the initial vertical spread of the fire from apartment 16 (which took more than 20 minutes), the fire is contained between two vertical columns.
Hypothesis 5: The construction of the vertical columns, which included cavity barriers and aluminium mounting and fixing profiles, played an important role in limiting horizontal flame spread, and this channels the vertical propagation during this period, ie, before the fire reached the architectural crown of the tower.
Observation 6: The rate of flame spread at the crown of the tower is the same for both the clockwise and anticlockwise spread.
Hypothesis 6: Wind did not influence the fire at crown level and during horizontal propagation phase.
Observation 7: There are no significant irregularities in the horizontal propagation across all floors analysed.
Hypothesis 7a: Individual elements of the crown, such as columns or corners, had no effect on horizontal flame spread. It is obvious that the integrity of the cavities were compromised when the ACM-PE external cladding had burnt away, irrespective of the integrity of the cavity barriers. The behaviour of the cavity barriers needs further experimental analysis.

| Mounting and fixing details
The mounting and fixing brackets for the aluminium cassettes (in conjunction with the cavity barriers), appears to have played a role in limiting initial lateral fire propagation (Hypothesis 5). Nevertheless, these elements were mainly made of aluminium and melted. Further investigations will be performed experimentally and numerically to understand their behaviour during the fire.

| CONCLUSIONS
In the aftermath of the Grenfell Tower fire, numerous studies have and will be produced trying to explain and/or to model the observed fire behaviour. In order to validate modelling of such fire, visual observations and quantitative data are of prime importance. This study draws together the numerous available data of the Grenfell Tower fire and presents their analysis and the additional observations that can consequently be made.
The initial vertical fire spread can be split into three phases: first, a linear phase is followed by a stagnation phase. The first phase is characterized by a constant spread rate of about 3.6 m/min. The stagnation phase is unusual and need to be explained. Then, before reaching the crown, power four behaviour is observed with a maximum vertical spread rate of 8 m/min when reaching crown.
Horizontal fire spread is quickest at the crown (0.293 ± 0.005 m/min). A linear correlation between the rate of horizontal fire spread and height is also obtained. This variation explains the rate of formation of V-shape induced by the propagation.
Construction details of the façade seem of prime importance for reconstruction and need to be studied further through modelling and testing. This includes windows failures, as well as cavity barriers, and mounting and fixing details.
Based on these and other observations, several hypotheses are proposed (see Table 4). These hypotheses will be the basis of reconstruction work performed by the authors.