An evaluation of a nuisance-emissions-discriminating smart mine fire sensor system was made in an operating coal mine. These field evaluations were conducted to determine the sensor system's ability to discern nuisance emissions, such as diesel exhaust, emissions from flame cutting and welding operations or hydrogen gas from a charging station, from real fires and to compare the number of falsely reported fire alarms generated between the sensor system and a standard carbon monoxide (CO) monitor. The sensor system's ability to operate successfully in the working environment of an operating coal mine was also evaluated. The smart mine fire sensor system consisted of four sensors whose data outputs were fused with the use of a neural-network-type computer program. Long-term trials were conducted in a haulage way, a belt entry, and a track entry. The system functioned successfully in the belt entry, in accordance with its developmental goals, where the sensor system even discriminated events not anticipated during development. It was not totally effective in the haulage way and track entry, though, due to a combination of significant diurnal air temperature variations, dust, and mechanically induced vibrations. Also, deteriorating rib conditions contributed to operational problems in the haulage way evaluation. In general, the smart mine fire sensor provided nuisance emissions discrimination and was shown to be a viable new approach for mine atmospheric monitoring, enhancing miner safety. This paper describes the in-mine evaluation of the smart mine fire sensor system and discusses recommendations for improving the system.