Intelligent mine gas detector system based on pattern recognition

Intelligent mine gas detector system based on pattern recognition is developed with three metal oxide semiconductor and a catalytic combustion sensors. The identifying system is carried out by microcontroller unit, a pulse width modulation circuit is adopted to heat the sensors, and an analog-to-digital converter is designed to preprocess the voltage signals, the setup of the detector function and alarm values is implemented by the keyboard, the concentration of the gas and the temperature are showed on the liquid crystal display screen. In this research, pattern recognition technique based on back-propagation neural network was applied for testing and discriminating gases. The network was successfully trained and can recognize unknown mixed gases of various concentration and compositions. Once it is well-trained, the computed weights are inserted into the system from keyboard and then it can be employed anywhere even without computer so as to realize offline operation. Besides, it also has such merits as high accuracy, being portable and convenient, low-cost and good explosion preventing capacity.