Mine Fire Source Discrimination Using Fire Sensors and Neural Network Analysis

Fire experiments were conducted in the Safety Research Coal Mine (SRCM) at the National Institute for Occupational Safety and Health, Pittsburgh Research Laboratory, with coal, diesel-fuel, electricalcable, conveyor-belt, and metal-cutting fire sources to determine the response of fire sensors to productsof-combustion (POC). Metal oxide semiconductor (MOS) and smoke fire sensors demonstrated an earlier fire detection capability than a CO sensor. This capability was of particular significance for a conveyor-belt fire in which the optical visibility was reduced to 1.52 m with an increase in CO of less than 2 ppm at a distance of 148 m from the fire. Application of a neural-network program to the sensor responses from each type of fire source resulted in correct classifications of coal, diesel-fuel, cable, belt, and metal-cutting combustion with a mean of 96% of the test data correctly classified. Introduction Fire detection in underground coal mines is important for early fire location and safe miner evacuation. Fire detection is aided by the in-mine ventilation which transports the fire products-of-combustion (POC) from the fire source to fire sensors, and it is impeded by the diluting effects of the ventilation which reduces the measurable signal. Early mine fire detection experiments in both normally ventilated and near zero airflow mine entries have been previously investigated experimentally [1, 2]. The results of that research showed the improved performance of ionization and optical smoke fire sensors over CO sensors. The next advancement beyond early detection, in addition to determination of the fire location, is the determination of the material burning, the mode of combustion, and the extent of fire growth. This knowledge can be applied to the determination of appropriate actions to be taken to extinguish a mine fire and to initiate escape and rescue procedures for miners. Various in-mine materials can provide the initial fuel for a mine fire. The common fire source materials considered in this program were coal, diesel fuel, electrical cable, conveyor belt, and acetylene gas used for metal cutting. The solid material combustion was advanced using heaters through a smoldering combustion stage to produce a slowly increasing range of measurable POC. The POC include CO and other oxidizable gases and smoke particles with submicron diameters. It is the types of combustion products and their rate of change which will indicate the combustion material, mode, and growth rate. One method which can be deployed to make these determinations is the use of multiple fire sensors to discriminate the POC. The rapid analysis of mine fire products with a neural network program has been reported elsewhere [3, 4]. These approaches utilize temperature and gaseous POC from combustion of home materials [3] and laboratory heating of coal [4]. CO and smoke sensors are used for the early detection of underground coal mine fires. While CO sensors, in the absence of cross-interference from other gases, respond to the concentration of CO, the response of smoke sensors in mining applications is presented in terms of the smoke optical density. Ultimately, the optical density depends upon the smoke mass concentration, smoke particle diameter, and the dielectric constant of the smoke particles. Ionization smoke sensors are more responsive to smoke from flaming combustion, and optical smoke sensors are more responsive to smoke from smoldering combustion. The intense turbulent combustion during the flaming stage reduces the average particle size associated with smoldering combustion. In addition to these fire sensors, there are MOS sensors