Experimental investigation of human tenability and sprinkler protection in hospital room fires

The effectiveness of sprinklers in protecting hospital patients in the room of fire origin was investigated by 14 experiments with a residential‐grade sprinkler system, and by two free‐burns. The fire load was UL 1626 corner fire. Tenability conditions were evaluated for the UL 1626 corner fire using gas temperature and species concentration measurements, and by calculating the fractional effective dose (FED) and fractional irritant concentration (FIC) with the comprehensive model of Purser and a more simplified method of ISO 13571. In the sprinklered tests, the average FED at 15 minutes was 0.8 ± 1 with 95% confidence, when using the Purser's method, and 0.2 ± 0.2 with ISO 13571. The difference was mainly caused by the assumption in the Purser's method that all NOx gases behave like NO2. Ignoring the NO contributions decreased the Purser's FED values very close to those of ISO 13571. In nonsprinklered tests, the FED and FIC values indicated definite incapacitation and possibly death 3 minutes after ignition. The sprinklers effectively increase the possibility of surviving, but the toxic effects may still be dangerous. In hospital and health care environments, many of the exposed persons may have lower‐than‐average tolerance.


| INTRODUCTION
Water sprinklers are commonly used to improve the fire safety in spaces where the early suppression by people is not guaranteed to occur. In health care units, water sprinklers can be used to protect patient rooms, common spaces, and auxiliary spaces. Primary response to a fire in a patient room is to evacuate the people from the room of fire origin, but it is often questionable if the health care personnel can perform the task without proper training and equipment. It may be possible that the fire service eventually evacuates the room, and the effectiveness of sprinklers in protecting the patients inside the room becomes then in question.
The fire protection performance of sprinklers has been widely investigated, and they have been found to be effective in cooling the room of fire origin and restricting the fire spread. 1 It is commonly understood that the major threat to occupants is caused by hazardous gases. The effects of different asphyxiant and irritating gases have been studied extensively, and the mechanisms of the major asphyxiant gases are well understood. [2][3][4][5] Two widely used engineering methods for toxicity assessment are the models of ISO 13571 6 and the model by Purser 3 taking into account wider range of gases than the ISO standard.
Previously, several full-scale fire tests studying smoke gas toxicity and tenability have been performed, both without and with sprinklers.
For example, Blomqvist et al 7 used Fourier transform infrared (FTIR) analysis to measure CO, HCl, and HCN concentrations in a reconstruction of a hospital fire that occurred in Sweden in the 1990's. The hospital and the reconstruction had no sprinklers. The results showed that after a smoldering phase of ca. 500 seconds, all the measured gas concentrations increased rapidly and reached their maxima at ca. 600 seconds. After that, the concentrations decreased quickly so that at ca. 700 seconds, they resembled the pre-flashover state. More recently, Guillaume et al 8 made a tenability assessment of nonsprinklered bedroom fires. It was concluded that the smoke alarms activated before the tenability was compromised. The analysis was done by using method described in ISO 13571 standard. Based on the gas analysis, nitric oxide (NO) was determined to be most important irritant. 8 The effect of the sprinklers on toxicity has not been studied very recently. O'Neill et al 9 have investigated the effect of sprinklers to toxic yields in the 1980's, concluding that sprinklers prevented flashover and cooled the room, but the hazardous threshold for carbon monoxide were exceeded at the test area. In the 1990's, Hietaniemi et al 10 studied the effect of water suppression on toxic yields in the small scale using controlled-atmosphere cone calorimeter with water spray inlet. It was concluded that water suppression can even double the yields of CO and HCN. However, even though the yield in g/g increases, water suppression can reduce the total yield in g, since the mass of burned material decreases, which has a positive effect on the safety of occupants. Shelley et al 11 have performed tenability analysis of TV set fires in a sprinkler protected compartment, inspecting survivability in post-sprinkler activation environment. fractional effective dose (FED) method by Purser was used to assess the tenability conditions. It was deemed that FED < 0.1 should allow for safe escape of nearly all exposed individuals.
In Finland, residential sprinkler systems are sometimes used in health care buildings. These systems are then classified according to the Underwriters Laboratories (UL) test standard UL 1626. 12 In this work, we investigated the efficiency of a real, installed sprinkler system in suppressing or controlling the UL 1626 corner fire scenario in hospital rooms. The assessment was done by measuring thermal and toxic conditions during 15 minutes fires. With gas concentration measurements and FED and fractional irritant concentration (FIC) indices we try to conclude whether the conditions are compromising tenability. Both Purser's and ISO 13571 models are used, and their differences discussed.

| Building and sprinkler system
The experiments were performed in a 1960's health care center facility of Sysmä, Finland. The building was taken out of use 2 weeks before the experiments. Fires were burnt in 14 different patient rooms and two storage rooms, having the ceiling height of 2.8 m and varying between 16 and 21 m 2 in floor area. The walls between the rooms and the horizontal slabs were concrete, but inside the room there were some light-weight structures, such as closets. These structures did not participate in fires. The rooms were connected to centralized supply and exhaust ventilation ducts with an air handling unit serving about 20 rooms. In seven sprinklered experiments, the ventilation ducts of the fire room were closed to investigate the effect of air exchange on tenability and the role of ventilation network in smoke spread to neighboring rooms.
The building had been retrofitted with a wet sprinkler system 10 years before the experiments. The sprinkler system was designed according to standard SFS 5980 which is normally used for residential buildings. Each room had two horizontal, wall mounted sprinkler nozzles (Tyco 1334, K = 60.5 L/min/bar 1/2 , T act = 68 C and RTI = 35 ms 1/2 ).
The system was inspected just before the experimental campaign. The pipe pressure in the vicinity of the test rooms was measured continuously. Before activation the pressure was 5.7 ± 0.2 bar, and after the activation of one nozzle, it was 2.7-2.8 bar which, according to the manufacturer's data, corresponds to a horizontal throw of 6.1 m and a flow rate of 100 L/min from one nozzle. As a result, 1.4 m 3 of water was poured to the room during each experiment. For the water management, holes were drilled to the floor to lead the water to the collecting system one floor below.

| Fire load
The fire load was as close to the UL 1626 living room scenario 12  According to the full-scale laboratory measurements by UL, 13 the heat release rate (HRR) of UL 1626 corner fire scenario is initially about 100 kW, increasing in t 2 -manner to 300-500 kW at 60 seconds, and reaching a level of 1500 kW in 80-95 seconds.
Experiments with UL 1626 fire load were repeated 14 times with the sprinkler system and twice with sprinkler system closed (freeburn). The corner of the fire inside the room was chosen randomly to cover the possible orientations and distances to sprinkler nozzles. In six of the sprinklered tests, the sprinkler nozzles were at the wall next to the corner of the fire. The horizontal distance from the wood crib to the nearest nozzle was then in the range 0.8-1.4 m. In the rest of the experiments, the sprinklers were on the opposite wall relative to the corner of the fire, and the distance to the nearest nozzle was in the range 3.2-3.8 m.

| Test procedure
Each test lasted 15 minutes, approximation of the average time in Finland that it takes from fire department to arrive at the scene and start an effective operation. Before each test, the sprinkler system was initialized to the city water system pressure. A fireman with breathing apparatus went inside the room, the door was closed, and the fireman ignited the pool and fabric strips using a torch. Measurements were started about 1 minute before ignition. The fireman stayed inside the room for the entire experiment, delivering observations through radio.
After 15 minutes of fire test, the sprinkler system water source was closed, remaining HRR was manually extinguished, and the smoke ventilated through an open window.

| TOXICITY ANALYSIS
Gases have two major ways of affecting people, by asphyxia or by irritation. The required exposure times and incapacitating concentrations of asphyxiant gases are significantly smaller than those for irritants.
Thus, asphyxiant gases have more potential to incapacitate humans.
Irritants, however, can cause inflammation in lung tissue and thus be lethal hours or even days after the initial exposure. 4 where t is time, subscript I refers to incapacitation as a toxic endpoint, and In above, X i is the concentration of gas i at given time in ppm (except for O 2 which is expressed in volume %),˙V is the volumetric flow of breathing (l/min), assumed 8.5 L/min for a person at rest and 25 L/min for light activity. NO and NO 2 are assumed to protect from cyanide poisoning, and thus X CN = X HCN − X NO − X NO2 . D is the assumed incapacitating level of COHb% in blood, being 30% for light activity and 40% for rest. Presented formula for V CO2 is a correlation that describes hyperventilation, that is, caused by carbon dioxide.
Denominator 7.1 (l/min) in Equation (5) is a suggested value for the respiratory minute volume of resting person at background CO 2 concentration. The effect of irritants is consider in the FED calculation with a factor called fractional lethal dose (FLD) that calculates a sum where FLD i are lethal doses (Table 1).
In ISO 13571, the FED calculation method only considers the asphyxiant effects of CO and HCN: where Δt is a time increment between measurement time instances, and In the ISO 13571 calculation method, the person is assumed to be in a light work.  Table 1). The formula for FIC is: The incapacitation concentrations in the two methods are of similar magnitude in general, but the incapacitating concentration for CHOH is eight times higher in ISO 13571 than in the Purser's method.
The interpretation of the FIC/FEC calculations is also slightly different in two methods: Purser suggests that the FIC values exceeding 1.0 will significantly impair the escape efficiency of occupants and the FIC values exceeding 5.0 will cause incapacitation in 50% of the exposed population. ISO 13571 suggests that FEC exceeding 1.0 will cause incapacitation in 50% of population. This is a major difference and will significantly affect the results as the incapacitation concentrations of the individual gases are of the same magnitude.

| Thermal environment
Temperatures of the thermocouple tree were averaged over the sprinkler tests, and separately over the two free-burns ( confidence limit for the temperature at the highest location. After sprinkler activation, temperatures decrease rapidly, even though the uncertainty of wet thermocouples must be kept in mind. In the freeburns, the temperature at the patient level exceeded 200 C momentarily, which would cause a thermal hazard to occupants. To understand the reason between the different FED results, we calculated the contribution of each gas to the outcome of Equation (1) in the end of the tests. Figure 8 shows these contributions for sprinkler tests and free-burns. In the sprinkler tests, NO x gases cause 71% of FED when the person is assumed to be at light work, and the contribution would be even higher for a person at rest. CO is the second most important gas, and the irritants (FLD) and O 2 depletion follow with similar contributions. ISO standard ignores NO x as asphyxiant gases, which leads to much lower FED values in Figure 7. In freeburns, HCN clearly dominates with 95% contribution, reflecting the dependence of gas formation mechanisms on fire temperature.

| Gas concentrations
The averaged results of the FIC calculations using both Purser's and ISO 13571 methods are presented in Figure 9. The ISO method shows significantly lower FIC values than the Purser's method because it assumes much higher irritant concentration for formaldehyde (see Table 1), which is present in 8 ppm average peak concentration. In sprinkler tests, the average FIC values remain below 0.3 but F I G U R E 6 FED results for each sprinkler test using the 'Purser'-method. 3

| DISCUSSION
The current FED models assume that all nitrogen oxides (NO x ) behave like nitrogen dioxide (NO 2 ), despite the fact that these two gases have very different toxic potencies and toxic mechanisms, NO 2 being clearly more toxic than NO. 15 The justification is that although nitrogen oxides are initially formed as NO, they gradually oxidize into NO 2 .
To investigate the sensitivity of our results to this assumption, we rec- standard, the probability of incapacitation can be related to FED using a log-normal distribution with median at FED = 1 and SD of 1.
The irritancy assessment using FIC leads to similar conclusions.
According to the Purser's method, the escape impairment level FIC = 1.0 might be exceeded in some sprinklered fires, but the incapacitating levels were only reached in nonsprinklered fires. If we assume that the effects of FIC also follow the curve shown in Figure 11, the conditions of all fires would be considered unsafe for the weakest individuals. More work is needed for the estimation of the actual survivability probabilities.

| CONCLUSIONS
Fire experiments with UL 1626 fire source were performed in a real health care center equipped with a residential sprinkler system. In addition to the gas temperatures and toxic gas concentrations, we  known to be very different, combining them in the FED calculation seems to be a strongly conservative approximation. Finding and assigning the lethal, incapacitating and escape impairment doses for these gases individually is essential for the accuracy of fire toxicity analyses.