Indoor Fire Prevention Based on Miniaturized Sensor Drones and Stationary Sensor Nodes †

: Miniature drones for indoor use and stationary sensor nodes have been equipped with thermal imaging and sensors for temperature, humidity, and relevant gas components to enable preventive analysis of indoor environments in order to identify pre-fire events.


Introduction
Prevention of fires and early fire detection in historical buildings has recently gained focus, e.g., due to the heavy damages caused by the fire at the cathedral of Notre-Dame in Paris.Fire events in heritage buildings often occur in places that are difficult to access and not covered by fire detection systems.Therefore, this work addresses modular stationary and autonomous sensor systems applicable in cultural heritage areas with limited options for installation and limited accessibility.
Drone systems equipped with various sensing capabilities are widely used for monitoring in outdoor environments [1].These drone systems are not suitable for indoor use due to their size and weight, which are in the range of several kg.Since the overall weight of drones for autonomous operation indoors must not exceed 250 g due to EU and German regulations, there is a need for further miniaturization of sensor drones for indoor use.In this work, the extension of miniaturized drone systems with additional sensing capabilities for the detection of pre-fire events in autonomous operation is described.

Materials and Methods
Based on the investigation and testing of different drone platforms, the Tello Edu Framework has been selected for the extension towards an autonomous sensor drone system.In addition to the camera and the sensors already existing in the drone setup to enable basic flight capabilities, the following sensors dedicated to identifying parameters indicating pre-fire events have been integrated: -Thermal Imaging: Melexis MLX90640, FLIR Lepton 3.5 -Multi-gas detection: Sciosense ENS 160, Sciosense ENS 210, Sensirion SCD30.
A sensor mount has been implemented allowing for easy sensors changes.To allow autonomous flight operation, a multi-ranger board has been adapted for obstacle avoidance.The actual version of the sensor drone platform is shown in Figure 1a.The overall weight is in the range < 200 g, depending on the sensors installed.The operation of the drone is controlled by a Robot Operating System (ROS)-based system.A mobile landing and charging platform has been created to allow autonomous operation (Figure 1b).

Discussion
The flight performance and sensing capabilities have been evaluated in test flights in laboratory environments and in a fire test laboratory.Figure 2a shows the response of the Sciosense ENS160 sensor to VOC (volatile organic compounds) arising from an open flame mapped to the position of the drone during a test flight.A steep increase in VOC sensor signal is observed if the drone reaches the area the fire event allowing to detect and localize the fire event.

Discussion
The flight performance and sensing capabilities have been evaluated in test flights in laboratory environments and in a fire test laboratory.Figure 2a shows the response of the Sciosense ENS160 sensor to VOC (volatile organic compounds) arising from an open flame mapped to the position of the drone during a test flight.A steep increase in VOC sensor signal is observed if the drone reaches the area of the fire event allowing to detect and localize the fire event.Figure 2b

Discussion
The flight performance and sensing capabilities have been evaluated in test flights in laboratory environments and in a fire test laboratory.Figure 2a shows the response of the Sciosense ENS160 sensor to VOC (volatile organic compounds) arising from an open flame mapped to the position of the drone during a test flight.A steep increase in VOC sensor signal is observed if the drone reaches the area of the fire event allowing to detect and localize the fire event.Figure 2b

Figure 1 .
Figure 1.(a) Drone equipped with Sciosense ENS160/210 multi-gas sensor and the multiranger board for obstacle avoidance; (b) mobile drone landing platform with integrated charging.
Figure 2b represents the thermal image of an hot object in the test room during flight with is classified by a threshold based algorithm as an possible risk for fire.Evaluation of additional data recorded during laboratory and fire tests combining sensor drone and stationary sensors shows the potential of miniaturized sensor drones to reduce risk of fire in endangered indoor areas.

Figure 1 .
Figure 1.(a) Drone platform equipped with Sciosense ENS160/210 multi-gas sensor and the multiranger board for obstacle avoidance; (b) mobile drone landing platform with integrated charging.
represents the thermal image of an hot object in the test room during flight with is classified by a threshold based algorithm as an possible risk for fire.Evaluation of additional data recorded during laboratory and fire tests combining sensor drone and stationary sensors shows the potential of miniaturized sensor drones to reduce risk of fire in endangered indoor areas.

Figure 1 .
Figure 1.(a) Drone platform equipped with Sciosense ENS160/210 multi-gas sensor and the multiranger board for obstacle avoidance; (b) mobile drone landing platform with integrated charging.
represents the thermal image of an hot object in the test room during flight with is classified by a threshold based algorithm as an possible risk for fire.Evaluation of additional data recorded during laboratory and fire tests combining sensor drone and stationary sensors shows the potential of miniaturized sensor drones to reduce risk of fire in endangered indoor areas.

Figure 2 .
Figure 2. (a) Response of the ENS160 VOC sensor to an open flame during flight; (b) hot object detection by thermal image recorded during flight.