Reconstruction of the Grenfell Tower fire—Part 5: Contribution to the understanding of the tenability conditions inside the apartments following the façade fire

The serious fire incident at Grenfell Tower in London, involving a combustible façade system that was installed as part of a major refurbishment of the building, has raised concerns regarding the fire risk that these systems pose. The fire spread over the façade of the Tower was previously numerically modelled and this model was validated by comparison with observational data. This model was used to determine the fire behavior of the façade and the fire's propagation into apartments through windows. In the present paper, impact models are used to evaluate tenability conditions inside the Tower, especially for the apartments in the first corner of the Tower that caught fire. The source of toxic effluents includes the components used in the refurbishment of the façade and the apartment furniture. Different hypotheses of gas yields are tested to assess variability and unknowns in the burning conditions. An extensive literature review was conducted to investigate the toxic yields to be considered in the simulations. Tenability conditions are assessed for each apartment during the fire spread over the façade. This leads to the quantification of the thermal and toxic environment inside the apartments. Two different models are tested for thermal and toxic threats, and the influence of the insulation material used in the façade is investigated. The results showed that the same conclusion can be made regardless of the input data for toxicity and the model used, within the limits of the studied dataset and conditions. Fires from the apartments quickly drive tenability conditions, independently of the dataset and model used, and even if mineral wool is used instead of poly‐isocyanurate as façade insulant.


Summary
The serious fire incident at Grenfell Tower in London, involving a combustible façade system that was installed as part of a major refurbishment of the building, has raised concerns regarding the fire risk that these systems pose. The fire spread over the façade of the Tower was previously numerically modelled and this model was validated by comparison with observational data. This model was used to determine the fire behavior of the façade and the fire's propagation into apartments through windows. In the present paper, impact models are used to evaluate tenability conditions inside the Tower, especially for the apartments in the first corner of the Tower that caught fire. The source of toxic effluents includes the components used in the refurbishment of the façade and the apartment furniture. Different hypotheses of gas yields are tested to assess variability and unknowns in the burning conditions. An extensive literature review was conducted to investigate the toxic yields to be considered in the simulations. Tenability conditions are assessed for each apartment during the fire spread over the façade. This leads to the quantification of the thermal and toxic environment inside the apartments. Two different models are tested for thermal and toxic threats, and the influence of the insulation material used in the façade is investigated. The results showed that the same conclusion can be made regardless of the input data for toxicity and the model used, within the limits of the studied dataset and conditions. Fires from the apartments quickly drive tenability conditions, independently of the dataset and model used, and even if mineral wool is used instead of poly-isocyanurate as façade insulant.  1 The fire spread to the façade via external flaming from an apartment located on a lower residential floor of the east face of the Tower. This has been extensively detailed in expert reports 2-5 and in video and photographic records of the real fire. These records were used to provide an analysis of the post-break-out vertical and horizontal fire propagation over the whole façade of the Grenfell Tower in reference 6.
The performance of the façade system installed on the Grenfell Tower was simulated using a model that was validated at intermediate and large scales, as addressed in references 7,8.The simulations closely matched the experimental results of reference 9 and confirmed that the aluminum composite material (ACM) cladding was the main element driving the global fire behaviour of the tested façade systems. In particular, systems that featured ACM cladding made with a polyethylene core (ACM-PE) showed extensive fire propagation regardless of the insulant used. 8,12 The fire development inside the initial apartment of Grenfell Tower and its behaviour at the kitchen window was investigated numerically in reference 10. The overall heat release rate (HRR) for typical apartment rooms and window failure criteria were estimated roughly, based on assumed apartment contents prior to the fire.
A complementary thermomechanical analysis of window failure was performed previously and reported in reference 11.
The full height of the Grenfell façade was modelled numerically using the computational fluid dynamics (CFD) code fire dynamics simulator (FDS) [14][15][16][17] to determine its fire behaviour. 12,13 The vertical and horizontal fire spread over the façade of the Tower were validated by comparison with video and photographic observations of the real fire.
The numerically predicted fire propagation was consistent with observations of the disaster. 6 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. 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. These flats are called "X1" to "X6" in this paper. For example, "X6" flats referred to the apartments from 16 (4th floor) to 206 (23th floor), and "X1" flats referred to the apartments from 11 (4th floor) to 201 (23th floor).
Observations from the fire, detailed in reports [2][3][4][5] and in reference 6, have shown that the spread of fire over the Tower can be split into different periods. During the period from 01:08 a.m. to 01:29 a.m., approximately, external flames spread over the east face of the Tower. Occupants of flats with windows located on that side of the Tower corresponding to the "X6" flats position (from flat 16 to flat 206, Figure 9), were the first that have seen flames close to their windows, followed by smoke and flames entering their flats. The abbreviation of the "X1" to "X6" localization is reminded in Figure 1. The fire originated in "X1" at the fourth level of the Tower. Occupants evacuated, and no fatalities were observed in "X6" flats. 3 The current paper presents an impact model that considers fire loads from the façade system and from apartment furniture. Tenability conditions are assessed in terms of thermal effects and toxicity [34][35][36][37] during the fire spread over the façade of the Tower. This allows the quantification of the conditions inside the Tower and an analysis of the contribution of the façade and other building fabric components, and that of the apartments' contents.
Existing methods often consider only smoke density and/or carbon monoxide, with yields often treated as constants, usually assuming a well-ventilated fire. This is the case for the CFD code FDS [14][15][16][17] used in this study, where the effluent yields are constant for a given combustion reaction, and will not depend on the ventilation condi- tions. An extensive bibliographic study was conducted to investigate toxic and asphyxiant effluent yields, mainly CO and HCN, to be used in simulations. The objective was to reproduce, in the simulations, the change in CO and HCN yields depending on the fire development. In the simulations, a change in the effluent released from the fire is assumed in the toxicity analysis, taking into consideration ventilation conditions. The different scenarios considered are detailed in

| SMOKE TOXICITY AND GRENFELL FIRE
A global synthesis of the previous research and main findings from Grenfell incident reconstruction is addressed in Figure 2. This multistep research was performed with highly interdependent parts, both experimental and numerical. The synoptic allows the understanding of the whole approach from the very first step of this research.
F I G U R E 2 Synthesis of the whole approach from the very first step of this research to the actual paper-highly interdependent parts, both experimental and numerical The Grenfell Tower fire resulted in 71 fatalities. Fire smoke reduces visibility and burns exposed skin, and also burns the mouth and nose if hot air is inhaled. The effects of heat exposure are dose-related and depend on the intensity of heat radiation or smoke temperature and on the duration of exposure. Smoke also contains irritants and asphyxiant gases. At the limit of tenability, irritant species will affect the eyes, nose and throat and cause breathing difficulties. They also affect behaviour to a certain degree, by inducing tears and coughing, and by limiting visibility and movement. 34 The effects of exposure to asphyxiants, including carbon monoxide (CO), hydrogen cyanide (HCN) and carbon dioxide (CO 2 ), depending on the inhaled dose over a period of time and thus, on the concentrations of these species and on the duration of the exposure. Their effects depend on the variability of human responses to toxicological injuries. Depleted oxygen may also be a parameter driving tenability. 18 Thus, the contribution of any burning materials in terms of mass loss rate, yields of combustion products, etc., as well as their burning conditions (well or under-ventilated conditions) must be considered to evaluate tenability in terms of toxicity and heat. However, a number of factors complicate the characterization of gases released from a fire. A fire is a dynamic and turbulent process and the concentration of specific compounds in the smoke may change from μl/L to percentage levels during the fire, or from one part of the plume to another. Thus, the composition of smoke gases is often very complex and changes rapidly with temperature and ventilation conditions. [19][20][21][22][23][24][25][26][27][28][29][30][31][32][33] For flat configurations such as those at Grenfell, the English Building Regulations, require compartmentation between flats and between flats and common areas. Thus, fire is supposed to be confined in the flat of origin and should not spread into the lobbies or, via the exterior façade, to other flats. The common lobbies should be separated by fire-resistant elements from the main escape stair, at least for 30 min in the event of a fully developed fire. Naturally, this value is highly theoretical and does not consider smoke leakage. However, these measures should prevent smoke from entering the lobbies and stairs, allowing occupants to evacuate in safe conditions without exposure to toxic smoke and heat.
During the Grenfell fire, different combustion ventilation regimes may have occurred. For the external façade system fire, the external cladding is thought to have burned in a well-ventilated regime because the system was directly supplied with oxygen from the exterior. The ventilation regime for the insulant is unknown and probably varied with time, especially being well-ventilated when the external cladding disappeared. For the apartment fires, the main fire source is provided by the furniture. The combustion is assumed to occur first in well-ventilated conditions, because of the ambient oxygen available, followed by a quick transition to an under-ventilated regime. This time scale of oxygen consumption is typical of compartment fires and was evaluated numerically in the previous research dealing with the assessment of representative apartment fires in Grenfell Tower. 10 In particular, the effluent yields for CO and HCN will increase when the oxygen concentration inside the apartment falls below 15%. Other effluents can be considered for toxic evaluation, such as hydrogen chloride (HCl) yield, which is not dependent on ventilation conditions. 32

| NUMERICAL SETUP
In previous work, 6    condition, window failure, heat release rate and effluent concentrations, are thus known at every location in the façade and inside the apartments (see Figure 2 in reference 13).
The evolution of window failure times, depending on the room involved, is evaluated numerically in reference 13 for the whole Tower.
It is compared with records of the presence of fire or smoke at a given floor 4 and allows the understanding of the fire propagation over the Tower. An overview of the window status is illustrated in Figure 3 However, window failure was probably not the only way of flame exit and re-entry, and the window failure mode is probably not unique during the Grenfell disaster as summarized in Table 1. The scenario leading to one of the earliest massive inflow of fire effluents appears to be the deformation of the window frame and is thus considered in the present work, although that especially when comparing smoke resulting from furniture and from facade materials, the gaps that may exist locally around the window has an influence before the window glass breaks.

| Fire threat models used
All models used and assumptions are valid for tenability and may differ from those used for lethality. Tenability assessment studies in fire safety engineering are based on an evaluation of the thermal effects and the effects of smoke toxicity. Tenability is defined as the ability of humans to perform cognitive and motor-skill functions at an acceptable level when exposed to a fire environment. It means that occupants in a fire situation should be able to escape by their own devices or to remain safe in specific locations, with acceptable physical and mental capabilities. The limit of tenability is characterized as incapacitation. ISO 13571 34 has been developed as a toolbox to perform tenability analysis. The standard proposes several models to determine the time to compromised tenability for thermal effects and toxicity, among other factors. In such analyses, the time to compromised tenability is the earliest time that is reached for an occupant. Tenability models are often questionable by nature, and this is why two different models have been used in this work: the original one from ISO 13571 and the evolution of it proposed by Pauluhn in 2018. 36,37 The models are used in this study to identify contributors to tenability more than to evaluate absolute tenability conditions in the apartments.
In the present paper, the objective is to use existing models and to compare their results to identify the degree of independence of the conclusion from very different models.

| Thermal models
Thermal effects may be due to the temperature of the air (convective effect) or to the received radiation (radiative effect). Two models are tested in the present publication.    39 It is presented in Equation (1) Thermal model #2 This model is an evolution of the previous one. 36  whelmed. This is highly dependent on ambient temperature as well as air humidity. Reference 36 does not recommend any value, but a temperature over 60 C for a few minutes in humid air, or 40 C for several hours, is considered sufficient.

| Toxic gas models
The concept of FED and fractional effective concentration (FEC) relate to the manifestation of physiological and behavioral effects in exposed subjects. Two models are used: the existing ISO 13571 model, adapted to fire safety engineering, and another model more adapted to a regulatory pass/fail approach.

Toxic gas model #1
This model is the original one from ISO 13571 in its 2012 revision. 34 Table 2. The uncertainty for FEC is estimated as ±50% in accordance with ISO 13571:2012.
Toxic gas model #2 This model has been developed recently. 36,37 As in toxic gas model #1, there are toxicity models for asphyxiants and irritants, as their physiological effect is different at the tenability level. Nevertheless, they are both based on parametrization using non-lethal threshold values, as recommended nowadays in inhalation toxicology, in alignment with REACH, AEGLs and other state-of-the-art protocols.
Some technical aspects are also different. Firstly, CO 2 is considered for its effect, not as an aggravating factor for hyperventilation as in toxic gas model #1. Secondly, irritants are also considered through their dose effect. The general aspect of the model is given in Equation (7), where C n i is the average incremental concentration, expressed in μl/L of the class-specific toxic gas I, AF is the assessment factor applied to calculate the incapacitation threshold based on the respective non-lethal threshold concentration, Δt is the chosen time increment, expressed in minutes and k i is the effect-based toxic load constant. For asphyxiants, the model considers the effect of CO, CO 2 and HCN as in Equation (8). For irritants, the parametrization of the equation is given by Table 3, taken from SLOT-DTLs values. 41 One of the major parameters governing the proposed model is the assessment factor (AF) used by the expert. An AF of 3 has been used for HCN and CO, and an AF of 10 for all irritants, including interspecies difference, intraspecies effects, and the different uncertainties sources. More details on AF are available in references 43-48.

| Significance of FED and FEC values
FED and/or FEC for both model #1 have been established by a median analysis of effects on the population. As a consequence, assessment of any exposed population supposes that the sensitivity across the population is taken into account, even roughly. Without further knowledge, assessment of the population supposes a lognormal repartition of the effects through the population. In other words, the significance of FED and/or FEC = 1 is that the effect would occur for 50% of the normal population. The significance of FED and/or FEC values in terms of coverage of the population is given in Table 4.
Such an approach is very useful for probabilistic studies but does not allow certainty of coverage of the population, and as a consequence is difficult for regulators to accept. FED for both model #2 is based on non-lethal endpoints. Contrary to the previous FED model for asphyxiants, parametrization indicates that the incapacitating effect of CO and HCN occur at one-third of the lethality endpoint. The effect of CO 2 is considered in its own right. For irritants, the model is dose-based so only cumulative.
Regarding the parametrization of the AF factor, it estimates that the incapacitation effect occurs at 1/10 of the lethality endpoint. The factor AF is selected by the user, but toxicological knowledge is required to make this selection properly. The coverage factor of such a model is very different from the previous one. Probabilistic assessment is lost for such a model based on pass/fail tenability criteria. Below FED = 1, no effect is supposed to occur in a normal population. Over FED = 1, effects start to occur. This kind of approach is ideal for a pass/fail assessment. It is used in inhalation toxicology and for many regulatory purposes, such as the REACH regulation in the EU, as it covers fully the general population.
Regarding the hypotheses presented, FED = 1 in model #2 is supposed to have approximately the same coverage factor as FED = 0.1 in model #1.

| Oxygen depletion model
Interpretation of effects due to dioxygen depletion is also included in the analysis. A threshold of 16% of dioxygen is considered as a concentration sufficient to compromise tenability in a few minutes. 42 The technical choice has been made to monitor effluents from the various products present in the façade, at windows reveals and inside the apartments, as presented in Table 5. All calculations also consider soot as essential in radiative transfer, and in thermal impact models.
The furniture contribution was estimated in a previous study 10 and considered the combustion of tables, chair, a carpet, sofas, a TV set, a wardrobe, a mattress, nightstands and electronic devices (a washing machine, a cooker a fridge freezer and a mini-fridge). The main gases that were considered are wood, polyethylene, polycarbonate and poly-urethanes. The main gases are hence CO 2 , CO and HCN.
Each quantity is evaluated in the middle of a room, at a height of 1.5 m, corresponding to typical mean nose and face height. The HCl yield of the PVC window reveals lining is not dependent on vitiation conditions and remains constant whatever the conditions are. 50 The yields used for this product in the study are 0.27 g/g for HCl, 0.063 g/g for CO and 0.176 g/g for soot. 10,12,49 The heat of combustion and heat release rate per unit area were provided in reference 49.

| Gas yields used for toxic threat assessment
All the combustion properties for the cladding material (ACM-PE), the window reveal insulant (PIR W) and the infill panels (XPS), in terms of CO, HCl and soot yields, the heat of combustion and heat release rate per unit area, are those used in previous studies of the Grenfell tower fire and detailed in associated references 7 , 8, 10, 12, 13.
A summary of the yields for the gas species considered is given in Table 6.
Initially  giving regard to the dimensions of the cassettes and the severity of the façade fire ( Figure 4).  Figure 5B). The delays correspond to the time difference between the ignition of the exposed face of the ACM-PE cladding (PE +), the unexposed face of the cladding (PEÀ), and the exposed face of the insulant (PIR+) as shown in Figure 5A). Below Floor 6, PE+ ignites before PEÀ and before PIR+. Therefore, the PIR insulant will be

| Combustion gas yield assumptions investigated
Three different combustion gas yield assumptions have been investigated for each of the apartment furniture and the PIR façade insulant.
A summary of the nine combined combustion gas yield scenarios is indicated in Table 9. The toxicity FEDs have been evaluated with the two toxicity models #1 and #2. The authors estimate scenario 1 as the most severe and scenario 6 as the central one, so the detailed results for these two scenarios are presented in the following sections, even though the nine scenarios have been calculated.

| Influence of mesh refinement
The simulation of the whole fire requires a coarse mesh to allow modelling to be completed on an acceptable timescale. So, before conducting (see Figure 11). This also allowed validation of the vitiation of this  The coarse mesh model seems to over predict the incoming flow from the exterior to the interior of the apartment, and thus that the effluent concentrations from the façade fire will therefore be overestimated in the analysis. The toxicity analysis presented in this paper in Section 5.3 was performed for the Fat 26 using the model ran with the fine and the coarse meshes. Few differences were observed so that confidence in the analysis for the higher flats is expected (Table 10).  Table 9. The oxygen concentrations are given in    Figure 14 for both toxic gas models. Thermal FEDs, according to thermal models #1 and #2, are also indicated.

| Application of the impact model for Flat 26 and Flat 196
The times at which tenability became compromised for the first criterion (thermal or toxic) are shown in Table 12. Relative differences between 1 and 10 min are observed between models #1 and #2, with tenability compromised first with model #2 except for the kitchen of Because the coarse mesh model seems to over predict the incoming flow from the exterior to the interior of the apartment, and thus that the effluent concentrations from the façade fire as discussed in Section 6.2, the toxicity analysis was performed for the Fat 26 using the model run with the fine and the coarse meshes. Few differences were observed as indicated in Figure 15 so confidence in the analysis for the higher flats is expected.

| Analysis of contributors to smoke toxicity inside the apartments
A summary of the times to reach under-ventilation and the times at which tenability became compromised for the first criterion (thermal or toxic) is shown in Table 13 for Flats 26, 96 and 196, for the two  (Table 9) using the two toxic gas models are given in Table 14 and shown in The effluents from the façade elements are mainly found in the "X6" flats because they are contained inside the initial vertical fire plume.
For comparison purposes, the relative influences of the main contributors to the toxic gas FED for gas yield scenario 6 are indicated in    for PIR and MW as observed in Figure 20C)) result in a higher incoming flow through the window as seen in Figure 20A Table 15. Figure 22 shows the total concentrations, and those for each source material, of the combustion products present in the kitchen and the living rooms of Flat 26, over time. In the kitchen, the CO is released mainly by the apartment furniture. The HCN is mainly released from the furniture and to a lesser extent by the PIR window reveal insulant. In the living room, the CO and HCN come mainly from the apartment furniture.
Similar observations are made for Flat 196 (Figure 22). In the kitchen, the CO is released mainly by the apartment furniture. The HCN is very low and is mainly released from the PIR reveal insulant.
In the living room, the CO is released mainly by the apartment furni-  Figure 23 for both toxic gas models. Thermal FEDs, according to thermal models #1 and #2, are also indicated.
The times at which tenability became compromised for the first criterion (thermal or toxic) are shown in   The main results of the present paper showed that the same conclusion can be made regardless of the input data for toxicity and the model used, within the limits of the studied dataset and conditions.
Fires from the apartments quickly drive tenability conditions, independently of the dataset and model used, and even if MW is used instead of PIR as façade insulant.
Even if the numerical model addressed in this paper correlates well with the observations during the Grenfell Tower fire, several modelling assumptions were needed and constitute limitations of the model. As discussed in, 10,12 the numerical hypothesis must be considered for the model developed for the accurate fine grid to be applied to a coarser one as used in the present model. Furthermore, the CFD code FDS used in this study considers constant effluent yields for a given combustion reaction, and these yields will not depend on the ventilation conditions. The main scenarios addressed in the toxicity analysis were based on an extensive literature review to investigate the toxic and asphyxiant effluent yields, to be used in calculations. The objective was to reproduce in the simulations the variation in the CO and HCN yields depending on the fire development.
The sensitivity analysis and uncertainty calculations of such work are very complex and will be included in further parts of the study.
The analysis of tenability conditions inside Grenfell Tower showed that the same conclusion can be made regardless of the input data for toxic gas yields and the model used, within the limits of the studied dataset and conditions. The overall conclusion is that fires involving furnishings in the apartments quickly drive tenability conditions, independently of the dataset and model used, and even if MW is used instead of PIR as a façade insulant.
Further steps of this research will be dedicated to the tenability evaluation in the lobbies and stairs of the Tower. Additional research will focus on the sensitivity analysis of all the previous steps of the reconstruction.