This article delves into the critical role of residential fire sprinklers in mitigating the hazards posed by e-scooter fires initiated by thermal runaway of lithium-ion batteries. Through a series of controlled experiments conducted in a purpose-built residential structure, the study examines the effectiveness of sprinkler systems in preventing flashover and controlling fire hazards. The experiments involve both closed bedroom and open living room scenarios, with detailed analysis of temperature, gas concentrations, and sprinkler activation times. The results demonstrate that residential sprinklers can effectively contain fires, prevent flashover, and maintain tenable conditions, even in the face of rapidly developing fires. The study also highlights the potential for burn injuries in close proximity to the fire before sprinkler activation. The findings underscore the importance of residential fire sprinklers in enhancing fire safety and protecting occupants from the dangers of e-scooter fires.
AI Generated
This summary of the content was generated with the help of AI.
Abstract
Lithium-ion battery (LiB) powered devices are in use every day and function as designed, however there have been reports of fatal fires involving LiB powered micro-mobility vehicles from around the globe. When a LiB fails and starts a fire, it is typically the result of the battery suffering thermal runaway. Failure of LIBs in micro-mobility vehicles has been shown to create a rapidly growing ignition source that is capable of igniting nearby combustible materials within the room of origin and forcing the room to flashover in less than a minute. The impact of such rapid-fire growth has been reported in the media around the world. This type of rapid-fire growth is almost unprecedented in residential buildings and calls into question: can residential sprinklers control such a rapidly developing fire? This study investigates the impact of such rapid-fire growth in a residential building with both full-scale bedroom and living room fires started from the thermal runaway of a LiB in a sit-on e-scooter. Results from this study quantify the impact of residential sprinklers and clearly show the effectiveness of residential fire sprinklers on fires resulting from thermal runaway of sit-on e-scooters.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
1 Introduction
Lithium-ion battery (LiB) powered devices have become an integral part of daily life in many parts of the world. While LiB-powered devices are in use every day and function as designed, there have been reports of fatal fires involving LiB-powered micro-mobility devices from around the globe [1‐6]. In 2023, New York City reported 267 fires, 150 injuries and 18 fatalities from fires started from LiB-powered micromobility devices [7]. The large number of micromobility fires and associated high casualty rate is believed to be due to the more than 65,000 app-based food couriers [8]. In response to these fire incidents, New York City passed legislation that requires battery-powered mobility devices and their batteries to be certified by an independent laboratory. In addition, the city launched a LiB safety education campaign and battery trade-in programs [7].
From a fire safety perspective, LiBs can experience thermal runaway, which may result in overheating, flaming, and in some cases, an explosion. Thermal runaway is a process where the LiB cell enters a cycle of uncontrolled self-heating, as the temperature of the battery increases, the number and rate of exothermic chemical reactions increases. These phenomena accelerate heat generation, which can exponentially increase the battery’s temperature. This produces an exponentially accelerating feedback loop. This feedback loop continues until the battery cell vents gases via a pressure relief vent or, in a potentially violent case rupture. When the cell ruptures, vented gases may ignite immediately or accumulate outside the cell and cause an explosion. Gases released from the cell include hydrogen, carbon monoxide, and a mixture of hydrocarbons, these gases are referred to as Thermal Runaway Effluent Gases (TREG) [9]. When battery cells are packed together, the thermal runaway process may spread from cell to cell until all the cells in the battery pack have been affected. Thermal runaway can be initiated by physical damage, overheating, and failure of the battery management system, which could result in overcharging or over discharging.
Advertisement
Figure 1 shows a sit-on e-scooter with a 60 V, 1.2 kWh battery pack made from 136 of the common 18650 LiB cylindrical cells soldered together with minimal insulation between the cells and the external metal case. In Fig. 1, the TREG was first visible at t = 0 s, the photo on the left was taken 12 s later, 1 s before ignition. The photo on the right was taken 1 second after ignition. The same model of scooter was used in a previous study in an unsprinklered house [10] and throughout this experimental study.
Fig. 1
Video images captured (1 s before and 1 s after ignition) from the experiment under the oxygen consumption calorimetry hood
An article in the New York Times provided background on two LiB initiated fires. In one case, the LiB from a sit-on e-scooter was charging in the living room of a 3rd-floor apartment when it exploded. The fire resulted in the death of an 8-year-old girl. In the other incident, an e-bike LiB was charging in a bedroom in an apartment. The owner of the e-bike saw smoke and then the LiB exploded. The owner was burned on their back and arms as they ran out of the apartment [11]. In addition to the generation of heat, the smoke production is another hazard from fire. Based on the news reports, the trend is clear that e-scooter fires resulting in fatalities and injuries tend to occur in residential occupancies. This raises the question of the capabilities of residential fire sprinkler systems to mitigate the hazard of a LiB-powered e-scooter thermal runaway fire.
It has been noted anecdotally in incident descriptions and documented quantitatively that e-scooter fires that started with the thermal runaway of a LiB result in a rapidly developing fire or possibly an explosion. Figure 2 compares the heat release rate (HRR) generated from two identical e-scooter experiments, as shown in Fig. 1, with that of an overstuffed upholstered chair [12]. The e-scooter fire has HRR between 1 and 1.5 MW in less than 20 seconds after the first signs of smoke from the LiB. In contrast, the upholstered chair, ignited with a small open flame, demonstrates a slower growth rate and had a peak HRR of approximately 1.8 MW, at 130 seconds after ignition.
Fig. 2
Heat release rate history for two e-scooters and an upholstered chair under an oxygen consumption calorimeter hood
Automatic residential fire sprinkler systems, designed and installed in accordance with NFPA 13D, Standard for the Installation of Sprinkler Systems in One- and Two-Family Dwellings and Manufactured Homes [13] have been shown to save lives and property. The purpose of the standard is “...to provide a sprinkler system that aids in the detection and control of residential fires and thus provides improved protection against injury and life loss.” In addition, a sprinkler system designed and installed in accordance with NFPA 13D is intended to prevent flashover in the sprinklered room of fire origin to improve the chance for occupants to escape or be evacuated [12]. According to the NFPA, the civilian death rate in homes with fire sprinklers was 89% lower than in homes without an automatic suppression system. A similar comparison was made on both civilian and firefighter injury rates in the homes with fire sprinklers, the injury rates were 31% and 48% lower respectively [14].
Advertisement
The NFPA 13D design criteria are based on a significant body of research [15] that guided the design requirements, including the design discharge of residential sprinklers, the minimum discharge density of 2.0 mm/min (0.05 gpm/ft2) [12], and the standardised tests included in UL 199 [16]. Fire suppression experiments demonstrated the importance of residential sprinklers delivering water high enough on the wall to cool the hot gas layer, prevent the fire from getting above the sprinkler discharge, and apply water to furnishings, such as beds or sofas, which may be located around the perimeter of the room [14].
Residential automatic fire sprinklers have been designed and are tested to suppress an upholstered furniture fire, while maintaining tenable temperatures in the test room [15]. With the rapid growth rate of a LiB initiated fire, the residential sprinkler system will likely be activated while the LiB is producing at or near its peak HRR. Residential fire sprinklers have been tested against dry Christmas tree fires, which also have a rapid-fire growth rate. In a study from NIST [17], the peak heat release rate of dry Frasier fir trees (moisture content of the needles was less than 10%) with an average height of 2.3 m, and an average mass of 12.7 kg, ranged from 3.2 to 4.3 MW. The peak HRRs were reached in 30 s or less after ignition. In the NIST sprinkler experiment, a dry Fraiser fir tree was positioned in the corner of a room approximately 3.5 m on each side and a ceiling height of 2.4 m. An open doorway, 0.9 m wide by 2.0 m high, was in the middle of the front wall of the compartment opposite the location of the tree in the back corner. The room was lined with painted gypsum board on the walls and ceiling and contained an upholstered sofa and chair, as well as carpeting, a table, and a lamp. When the tree was ignited in the furnished room without a sprinkler, the room transitioned through flashover within 60 s and the peak HRR exceeded 6 MW. In the experiment where a pendent style residential sprinkler was installed in the center of the compartment, the sprinkler activated 10 s after ignition and flowed from a water supply set to 34.1 L/min (9 gpm). The activation of the sprinkler prevented flashover and limited the peak HRR to approximately 1.8 MW at 60 s after ignition [16].
3 Experimental Setup
The objective of the experiments presented in this paper is to examine the effectiveness of a residential automatic fire sprinkler system, designed to comply with NFPA 13D, against the hazard presented by an e-scooter fire resulting from thermal runaway of the LiB module. Effectiveness will be evaluated by determining the ability of the system to mitigate the fire hazard and by the ability of the system to maintain tenable conditions in the experimental structure.
3.1 Experimental Structure
The experiments were conducted in a full scale, purpose-built, ranch style residential structure located on the grounds of the Delaware County Emergency Services Training Center (ESTC) in Sharon Hill, Pennsylvania. The experimental structure had a floor area of approximately 98 m2 (1 060 ft2) with 2.4 m (8 ft) tall ceilings. The structure was divided into four bedrooms, and an open design living room, dining, and kitchen area. A floor plan of the structure with major dimensions is shown in Fig. 3. Each window in the structure was 1.2 m high by 0.9 m wide with a sill height of 0.61 m. The windows had a vinyl frame and double-pane glass. The interior doors were composed of a textured, molded wood composite with a hollow core. The dimensions of the doors were 2.0 m high by 0.76 m wide. The exterior doors were stamped steel over a wood frame. The dimensions of the exterior doors were 2.0 m high by 0.90 m wide.
Fig. 3
Floor plan of the experimental structure showing the dimensions, sprinkler layout, and instrumentation plan
The walls were constructed from 38 mm by 88 mm wood studs spaced 0.41 m on center and filled with R-13 fiberglass insulation. The interior walls were covered with 12.7 mm thick gypsum board finished with latex paint. The exterior walls were covered with 6 mm thick fiber cement board siding, over a layer of olefin home wrap, and 11 mm thick oriented strand board (OSB) sheathing. The floors were built on 38 mm by 190 mm wood joists on a concrete slab. The floor joists were covered with 20 mm thick OSB. The gypsum board ceiling was fastened to the bottom member of the roof trusses which were spaced 0.41 m on center.
3.2 Sprinkler System
A residential sprinkler system designed and installed by the National Fire Sprinkler Association to meet NFPA 13D requirements was used for these experiments. A pre-plumbed pump with a 1.1 kW motor and a 1600 L (422 gal) tank system was used to supply water to the sprinklers. The pipe system was composed of nominal 25.4 mm CPVC piping. The water supply was located on the right side of the structure. Between the pump and the riser entering the structure was a turbine flow meter with a 19 to 190 L/min range. Once the riser pipe entered the attic space of the structure, a tee fitting allowed for two branch lines to be installed near the centerlines of the rooms along the front and the back of the structure. Sprinklers were installed near the center of each of the bedrooms and centered between the walls of the kitchen and living room. Figure 4 shows Bedroom 1 just prior to the experiment, showing the sprinkler head on the ceiling. Two sprinklers were also installed along the center line of the hallway and opposite the center of the doorway of Bedroom 1 and Bedroom 3, as shown in Fig. 3. Test drains that could flow to the exterior of the structure were installed at the end of each branch line. With an open sprinkler installed in each of the drain line outlets, the system supply was set to flow 100 lpm ± 2 lpm (26 gpm ± 0.5 gpm).
Fig 4
Photograph of the bedroom fire scenario showing the fuel load and e-scooter in situ, just prior to the experiment
The residential fire sprinklers used in these experiments were Reliable model R3516. These are fast response, standard coverage, pendent sprinklers with a 3 mm in diameter thermal element that has an activation temperature of 68 °C. The sprinkler has a K factor of 70.6 l/min/bar1/2 [18].
3.3 e-Scooter
The e-scooter was a sit-on type of two-wheeled scooter, with the battery module located under the seat. The scooter has a metal frame with plastic body panels, fenders, light lenses, and an instrument panel. The e-scooter battery had an advertised capacity of 60 V, 20 Ah, 1.2 kWh. With the battery installed, the e-scooters weighed approximately 56 kg. The battery module was composed of the batteries, cabling, and the battery management system, positioned inside a steel housing. Each battery module had an assembled mass of 9.0 kg. The steel housing has a mass of 1.92 kg. Figure 4 shows the e-scooter in place for one of the bedroom fire experiments.
In the living room scenario, the e-scooter is positioned just inside the front door along the entry path between the door and the hallway that leads to the bedrooms. The location of the e-scooter is shown on the floor plan in Fig. 3 and in the photograph of the living room in Fig. 5.
Fig. 5
Photograph of the living room with the fuel load in place for both scenarios, the scooter is just visible behind the sofa for the living room scenario
The fuel load consisted of the e-scooter and furnishings that would normally be found inside a residence. The furnishings were considered target fuels, in place to assess if they ignite or if the water from the sprinkler can protect them from the heat.
The bedroom furnishings, shown in Fig. 4, included a twin bed, two wood nightstands, two lamps, a dresser, curtains, carpet, and carpet padding. The twin bed included a box spring, mattress with polyurethane foam mattress topper, sheets, comforter, and a pillow. The total mass of the bedroom furnishings fuel package was approximately 106 kg.
The living room furnishings, shown in Fig. 5, included a three-seat sofa, a two-seat sofa, flat screen TV, TV stand, coffee table, two end tables, lamp, curtains, carpet and carpet padding. The sofas had a polyester, microfiber fabric covering. The seat cushions were filled with polyurethane foam cushions with a layer of polyester batting on the top and bottom. The back cushions were filled with polyester fiber filling. The total mass of the living room furnishings fuel package was approximately 140 kg.
The kitchen fuel load consisted of cabinets and countertops composed of engineered lumber, covered with plastic laminates, and vinyl flooring. The total mass of the kitchen fuel load was approximately 280 kg. For these experiments, the fire was started with the intentional generation of a thermal runaway condition in the e-scooter LiB.
3.5 Instrumentation
The house was instrumented to continuously measure gas temperatures, pressures, and gas species concentrations in the room of fire origin and adjacent areas. Figure 3 shows the general arrangement of the instrumentation used during these experiments.
Gas temperatures were measured with a 1.25 mm (0.05 in.) bare-bead, Chromel-Alumel (type K) thermocouple (TC) arrays and a single thermocouple (STC) installed adjacent to the sprinklers at the ceiling level. In the fire rooms, 1.6 mm (0.0625 in.) inconel-sheathed thermocouples were used in the thermocouple arrays. Each thermocouple array consisted of eight thermocouples with the top thermocouple in each array located 25 mm (1in.) below the ceiling and the remaining seven thermocouples spaced at 0.30 m (1 ft) intervals. Small-diameter thermocouples were used during these experiments to limit the impact of radiative heating and cooling. The total expanded uncertainty associated with the temperature measurements from these experiments was estimated to be ± 15%, as reported by researchers at NIST [19, 20].
A total of sixteen continuous gas concentration sampling ports were installed in the structure. In the hallway, the sampling ports were installed at 0.30 m (1 ft) and 0.91 m (3 ft) above the floor. In the other locations, the living room and bedrooms, the ports were installed at 3 ft and 5 ft above the floor. The locations of the ports can be seen in Fig. 3. Only 4 of the ports were used during each experiment, two were in the room of origin (RoO) in the corner opposite the e-scooter (labeled as Remote), and two were centered in the hallway, directly in front of the RoO door (labeled Hall-BRX, where the X is the Bedroom number).
Gas samples were analyzed using oxygen (paramagnetic alternating pressure) and a combination carbon monoxide/carbon dioxide (non-dispersive infrared) analyzer. The gas sampling instruments used throughout the series of experiments demonstrated a relatively expanded uncertainty of ± 1% when compared to span gas volume fractions [21]. Given the nonuniformities and movement of the fire gas environment and the limited set of sampling points in these experiments, an estimated uncertainty of ± 12 % was applied [22]. To minimize transport time through the system, samples were pulled from the structure using a vacuum/pressure diaphragm pump rated at 0.75 CFM. The sampling ports consisted of 9 mm (3/8 in.) OD stainless steel tubing within the structure. Once outside the structure, the sample was drawn through a condensing trap to remove moisture and filtered through a 5-micron polyester filter and a 3-micron polyester filter. At the exit of these filters, the sample line transitioned from stainless steel to polyethylene tubing until the sample reached the analyzer/pump rack. At the inlet to the rack, the gas flowed through a 0.3-micron HEPA filter before reaching the sample pump. Downstream of the pump, but upstream of the analyzer, the sample flowed through a desiccant (DRIERITE) filter [23] to remove any remaining moisture and finally a 0.01-micron filter. Before every experiment, the transport time of a known calibration gas from each sample port to each respective analyzer was measured. This time lag was accounted for in post-processing to ensure the gas data was in sync with the other measurements.
4 Results
4.1 Closed Bedroom Experiments
Four experiments were conducted in the bedrooms of the purpose-built structure, as shown in Figure 3. In the first two experiments (Bedroom 1 and 2) the battery was overcharged to cause thermal runaway. In the other two experiments (Bedroom 3 and 4) the battery was overheated with the electric resistance heater affixed to 5 of the cells within the battery pack. In all experiments, the battery was in a 100% state of charge (SOC). The time from the first visible signs of TREG from thermal runaway of the e-scooter battery, until the TREG ignites, and the room is vented, ranged from 12 to 93 s. This wide variance is likely due to the number and location of the cell(s) initially in thermal runaway and the propagation to the other cells within the battery pack. Following the ignition of the accumulated TREG cloud, the sprinklers activated in 3-5 s, containing the fire to the e-scooter seating area where the battery was located. Flashover in the space never occurred in any of the sprinklered fire experiments. The bedroom doors in all of the bedroom experiments were closed, and the natural swing was into the bedroom, which limited the failure to the exterior bedroom windows or the bedroom walls. The windows were double-pane, allowing a greater pressure rise than single-pane windows. In two of the bedroom experiments, the walls were displaced and vented the RoO. In one case, two nails pulled out of the stud-bottom plate connection, allowing the sole plate of an interior wall to move approximately 100 mm. In the other case, a portion of an exterior wall adjacent to a window frame was displaced. Both instances produced gaps between adjacent drywall sheets, which may have allowed the spread of fire into the wall if the fire had been sustained. In these sprinklered experiments, the fire did not spread into or through the gaps in the walls. Figure 6 shows a photo of the displaced wall after the stud detached from the sole plate.
Fig. 6
Photograph of the displaced wall as a result of the overpressure from the ignition of the gases released from the E-scooter battery
Table 1 provides the sprinkler water flow rates from each experiment based on the inline flow meter. NFPA 13D [12, 17] requires a minimum design discharge of 2.0 mm/min⋅m2 (0.05 gpm/ft2) or the sprinkler listing, whichever is greater, to the design sprinklers. The systems are designed to supply two sprinklers. The listing requires 64 lpm (17 gpm) for a coverage area of 5.5 m (18 ft) by 5.5 m (18 ft) (RASCO135) [17]. With the system supply set for a total flow from two sprinklers, the first experiment was conducted. This resulted in a flow rate through the single sprinkler in Bedroom 1 of 68 lpm (18 gpm). Given that the resulting design discharge was 5.6 mm/min (0.14 gpm/ft2), the system flow was reduced for the remaining bedroom experiments to provide a total flow rate of approximately 49 lpm (13 gpm), the listed flow rate for a 3.7 m (12 ft) by 3.7 m (12 ft) coverage area. This provided a nominal design discharge of 3.6 mm/min (0.09 gpm). The design discharge for each test is based on the measured flow rate and room area from Fig. 3 are given in column 4 of Table 1.
Table 1
Number of sprinkler heads activated and total water flow rates for each experiment
Room of origin
Number of sprinklers activated
System water flow rate lpm (gpm)
Design discharge mm/min (gpm/ft2)
Bedroom 1
1
68.1 (18.0)
5.8 (0.14)
Bedroom 2
1
48.5 (12.8)
3.5 (0.087)
Bedroom 4
1
49.6 (13.1)
4.2 (0.10)
Bedroom 3
1
51.5 (13.6)
3.8 (0.092)
Living Room
4
144 (38.0)
3.1 (0.076)
Figure 7 shows a sequence of 9 images captured from a floor level video camera in Bedroom 2 experiment, the time since ignition is shown in the lower right corner of each image. The first six images are at 1 s intervals up to sprinkler activation, and the last three images, showing sprinkler control, are at 5 s intervals. The first two images show the buildup of TREG before ignition. The following four images show the rapid-fire development from the e-scooter within the bedroom until the sprinklers activated. The final three images are 5 s apart after sprinkler activation as the sprinkler controlled the fire.
Fig. 7
Video sequence for the closed bedroom two scenario, the sequence starts 10 seconds after the first visible smoke and ends 30 s later when the sprinkler controls the fire
Figure 8 shows the history of the average upper layer temperature (average of the top 3 TC) and the average lower layer (average of the lowest 3 TC) on each TC array in the corner opposite the e-scooter in the RoO and the hallway immediately outside the RoO. Data was collected while the battery was being overcharged, for more than 4 hours in some cases. For the bedroom scenario, the time of most interest was the time interval starting 60 s before ignition of the TREG cloud until 180 s post ignition. The fast response of the sprinkler and rapid control of the fire can be seen in the data and provide the most detail in the figures. The sprinkler system was allowed to run for 10 minutes after activation, at which time the water flow was shut off. After the sprinkler system was shut off, conditions were monitored for any reappearance of fire. With no fire evident and no increase in temperature, firefighters entered the structure and removed the e-scooter.
Fig. 8
Average temperature for the three highest and three lowest thermocouples on the TC arrays in the corner opposite the e-scooter in the bedroom experiments
The vertical axis at t = 0 s has been highlighted in the graphs. The sprinklers activated between 3 and 5 s after the ignition of the TERG and this is indicated as the second dark line at t=5 s on all graphs. Temperatures from all four bedroom experiments are included in Figure 8 for comparison. In Bedroom 2, the ignition of the TREG results in a sudden temperature rise that peaks at 285 °C (545°F). In the other bedroom experiments, the initial ignition was not as intense, and the average temperature rise was less than 135 °C (275°F), and the lower layer temperatures were less than 70°C (158°F). The plume temperatures (not included in Figure 8) were more than 800 °C (1500°F) and would cause serious injury or death to anyone in close proximity to the scooter after ignition.
Immediately outside the door to the RoO, the upper layer temperature never exceeded 61 °C (142°F) in any of the experiments, so the temperatures in the hallway were not included in Fig. 8. This demonstrates that even a closed hollow core door can provide some level of protection to the occupants outside the RoO. The effectiveness of the closed door is supported by the successful “Close Before You Doze” [24] public safety campaign promoting sleeping with one’s bedroom doors closed. The other bedrooms, kitchen, and living room do not indicate a significant temperature rise for the first 180 s after observing smoke from the scooter.
Figure 9 shows the CO concentration history for the RoO in the corner opposite the e-scooter. All of the bedroom experiments are included in Fig. 9 for comparison. The maximum concentration and most aggressive growth rate in the CO concentration was seen in the bedroom 2 experiment, where the CO concentration rose rapidly following the explosion. Outside the RoO, there was only a minor increase, CO concentration <0.02, or 1% of the concentration in the RoO, so CO concentrations outside the RoO were not included in Fig. 9.
Fig. 9
CO concentrations in the RoO for the bedroom experiments, outside the bedroom door of origin, the CO concentration is < 1% of the RoO concentration and has not been included on the graph
When assessing the impact of the toxic smoke on the occupants, the Fractional Effective Dose (FED) is often used to evaluate the cumulative effect of the toxic products in the smoke and the occupants increased respiration rate caused by heightened CO2 concentrations. The procedure for calculating the FED is taken from Purser and McAllister’s chapter of the SFPE Handbook of Fire Protection Engineering [25]. An occupant’s actual FED depends on the escape route, travel speed, and local species concentration. The FED is shown in Fig. 10 is calculated in the RoO based on the available species concentrations in the corner remote from the e-scooter, see Fig. 3. The calculations can only use the species that were measured, i.e., carbon monoxide as the toxicant and CO2 causing increased respiration. No other toxic products were measured or included in the analysis. The ignition of the TREG occurs at t=0 s. The FED rises after ignition, corresponding to the increase in CO concentration in the RoO. Before ignition in some of the experiments, FED>0 indicates the toxicity of the TREG discharged in the RoO before ignition.
Fig. 10
Fractional effect dose for the closed-door bedroom scenarios calculated from the CO, CO2, and O2 measurements in each room
When assessing the impact of the FED on the building occupants, a Log-normal distribution, with a standard deviation equal to 1, can be used to characterize the human population as described in the appendix of ISO 13571 [26]. An FED=1 is considered to be fatal to 50% of the population. An FED = 2 would be considered fatal to 76% of the population, an FED=0.3 (a common FED used in fire safety design [24, 25, 27, 28] would be considered fatal to 11% and FED = 0.1 would be fatal to 1% of the population. Thus, an FED = 1 is considered to be fatal to the occupants. Looking at Fig. 10, the FED rises above 0.3 in as little as 9 s, and given the size of the room, an occupant may be able to escape. However, given the intensity of the TREG explosions, an occupant in such a small room would likely suffer significant burn injuries and be unable to escape. If an occupant was incapable of self-evacuation, they would likely be overcome by the CO within the RoO before someone could rescue them. In these experiments, the bedroom door to the hallway was closed and FED outside the RoO (results not shown) was less than 0.02 at 3 minutes post ignition, thus occupants that are outside the RoO would be expected to be able to self-evacuate before being overcome by the fire gases.
4.2 Living Room Experiment
For the living room experiment, the e-scooter was in a much larger space compared to the bedrooms with closed doors. The open living space that included the living room, kitchen, and hallway was more than 3 times larger than any of the bedroom experiments. In addition, the TREG ignited much earlier than the bedroom experiments, approximately 4 s after the TREG was first visible from the e-scooter. The shorter time to ignition cannot be easily explained due to the complexities of thermal runaway of LiB and the general stochastic nature of uncontrolled ignition. The consequence of the shorter ignition delay was less time for the TREG to build up, which resulted in only a 1 m diameter cloud around the e-scooter. Coupled with the larger space to disperse the pressure rise and the smaller explosion, there was no structural damage or window breakage.
For the living room scenario, the residential sprinkler system supply was restored to the system flow for supplying two sprinklers at 64 lpm (17 gpm). Table 1 includes the living room experiment, where four sprinklers activated, resulting in a system flow of 144 lpm (38.0 gpm). A hydraulic analysis was used to estimate the sprinkler flow in the living room experiment. The analysis divides the total flow as: 44.3 lpm (11.7 gpm) in the living room (closest to the e-scooter), 37.8 lpm (10.0 gpm) in the kitchen, and 30.9 lpm (8.5 gpm) from each of the two heads that activated in the hallway.
Figure 11 shows a sequence of 9 images captured from a floor level video camera; the first five images show up to sprinkler activation. Ignition of the 1 m diameter cloud is just visible in the second image. The following three photos show the large fire plume driven by the LiB that scorches the ceiling above the e-scooter. Five seconds after ignition, the sprinkler activates, controlling the fire. The final four images show the sprinklers controlling the fire over the next 120 s.
Fig. 11
Video sequence for the living room scenario, the sequence starts 6 seconds after the first visible smoke
Figure 12 shows close-up photos of the seat cushion on the synthetic sofa that was within 1 m of the e-scooter. The photos show the damage from the hot and burning material ejected from the LiB that landed on the sofa and melted deep holes through the fabric and into the padding material below the fabric. The inset image is a close-up photo that shows the deep penetration of the burning material into the cushion. The crater shown in the close-up image was more than 25 mm in diameter. Although the sofa was clearly under extreme attack from the ejected material from the burning e-scooter, the sprinklers were able to wet the sofa, thus containing the fire to the e-scooter.
Fig. 12
Close-up photos of the sofa, less than 1 m from the e-scooter, multiple burn marks can be seen in the sofa that could not ignite and spread the fire when the sprinkler wetted the surface
Figure 13 shows the history for the average upper and lower layer temperature calculated from the highest and lowest three thermocouples on each TC array. The TC arrays were in the living room corner, remote from the e-scooter, kitchen, mid hall, and end of hall. The ignition of the TREG gas cloud occurred at t = 0 s on the graph, and the sprinklers activated 5 s later, controlling the e-scooter fire. The larger volume and smaller TREG cloud resulted in a slower overall temperature rise compared to the closed bedroom experiments, even though the temperature rise is considered extremely rapid compared to a typical residential fire growth rate. The highest upper layer temperature is seen in the kitchen area. However, on closer inspection of the data, the higher upper layer temperatures are driven by the top two TCs (2.41 m and 2.13 m). At 1.83 m after 15 s, the temperature is less than 70 °C (158°F).
Fig. 13
Average temperature for the three highest thermocouples on the center TC arrays in each room along the escape path from each bedroom to the final exit
In this experiment, the shorting out of the e-scooter battery caused a great deal of interference on the thermocouple channels of the data acquisition system. This can be seen in the data shown in Fig. 13 from − 1 s to 31 s.
Figure 14 shows the CO concentrations history for the living room in the corner opposite the e-scooter and sampling position at the start of the hallway. Although the start of the hallway position is closer to the scooter, the momentum of the fire and smoke flow from the scooter was across the living room ceiling. This resulted in the CO concentrations in the living room increasing a few seconds before the increase of CO measured in the hallway. The peak values are similar to the CO peak values from bedroom experiments 1, 3, and 4. This seems to be related to the evolution of the smoke cloud and resulting fire from the e-scooter LiB.
Fig. 14
CO concentrations for the living room e-scooter scenario as measured in the living room and at the start of the hallway.
Figure 15 shows the FED history, calculated for each of the 4 four gas sampling locations, two in the living room and two at the start of the hallway. The FED 300 seconds after ignition of the TREG cloud was less than 0.2, indicating that a healthy occupant capable of self-evacuation would be able to evacuate.
Fig. 15
Fractional effect dose for the living room e-scooter scenario calculated from the CO, CO2, and O2 measurements in the remote corner of the living room.
Previous research [10] in a similar building geometry has shown that an identical e-scooter without sprinklers can take a room to flashover in less than 60 s. Figure 16 compares the upperlayer temperature (based on the average of the three highest thermocouples in the RoO array) for bedroom 2 and living room experiments discussed above, and the similar unsprinklered bedroom and living room experiments from [10]. In the unsprinklered case, the RoO reached flashover approximately 30 s after the ignition of the TREG, and external water was applied approximately 10 s later, causing the drop in temperature. Without intervention, the fire would have spread throughout the building. When sprinklers are present, they activate within seconds of ignition and control the fire by preventing flashover. This comparison of the results clearly shows that without sprinklers, an e-scooter undergoing thermal runaway in residential buildings is not survivable in the RoO and can be expected to threaten occupants beyond the RoO.
Fig. 16
Upper layer temperature (based on the highest three thermocouples in a vertical array) for e-scooter in a bedroom and living room fire scenarios with and without residential sprinklers
performed as expected in that it prevented flashover and prevented the fire from spreading to other fuels in the room after sprinkler activation. This was true even in the case of the living room, which resulted in the activation of four sprinklers.
Unfortunately for persons intimate with the fire, such as in the confined space of the bedrooms, there is a potential for burn injury in the seconds before sprinkler activation. The consumption of oxygen, combined with the increase in CO and CO2, provides the conditions for incapacitation or death if the person cannot quickly evacuate from the room.
In the living room fire, again a person in close proximity to the fire could be burned before the activation of the sprinkler. However, the larger open volume of the living room, kitchen, and hallway allows for the smoke to be disbursed and slows the onset of untenable conditions, which enables time to shelter or exit.
The residential automatic fire sprinkler system demonstrated the capability to control LiBs from a shielded battery module in a sit-on type e-scooter. Future research should be conducted to examine the effectiveness of automatic fire sprinklers on LiB fires in areas with residential battery energy storage systems or commercial applications like parking garages, battery storage areas, and recycling centers.
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Fleischmann C, Weinschenk C, Madrzykowski D, Schraiber A, Gaudet B (2025) Quantifying the fire hazard from Li-ion battery fires caused by thermal runaway in e-scooters. Fire Technol. https://doi.org/10.1007/s10694-025-01707-zCrossRef
National Fire Protection Association, Quincy, Massachusetts (2022) NFPA 13D, Standard for the installation of sprinkler systems in one- and two-family dwellings and manufactured homes
Fleming RP, Madrzykowski D (2008) NFPA fire protection handbook residential sprinkler systems, 20th edn. National Fire Protection Association, Quincy Massachusetts
16.
Underwriters Laboratories, Northbrook, Illinois. UL 199, Standard for Automatic Sprinklers for Fire-Protection Service, 13th ed., April 2024
Blevins LG (1999) Behavior of bare and aspirated thermocouples in compartment fires. In 33rd Proceedings National Heat Transfer Conference, pp 15–17
20.
Pitts WM, Braun E, Peacock R, Mitler H, Johnson E, Reneke P, Blevins LG (2003) Temperature uncertainties for bare-bead and aspirated thermocouple measurements in fire environments. ASTM Spec Tech Publ 1427:3–15
21.
Bundy M, Hamins A, Johnsson EL, Kim SC, Ko GH, Lenhart DB (2007) (2007) Measurements of heat and combustion products in reduced-scale ventilated-limited compartment fires, NIST technical note 1483. National Institute of Standards and Technology, Gaithersburg, MD
22.
Lock A, Bundy M, Johnsson EL, Hamins A, Ko GH, Hwang C, Fuss P, Harris R (2008) Experimental study of the effects of fuel type, fuel distribution, and vent size on full-scale under ventilated compartment fires in an ISO 9705 room, NIST Technical Note 1603. National Institute of Standards and Technology, Gaithersburg, MD
Purser DA, McAllister JL (2016) Assessment of hazards to occupants from smoke, toxic gases, and heat, chapter 63, SFPE handbook of fire protection engineering, 5th edn. Springer, New York
26.
ISO13571 (2012) Life-threatening components of fire - guidelines for the estimation of time to compromised tenability in fires, 2nd edition, International Standards Organization. https://www.iso.org/standard/56172.html
