Design Of Neural Net To Detect Early Stage Of Fire And Evaluation By Using Real Sensors' Data

Sensitive fire sensors are necessary to detect fires in their early stage, but they often produce false alarms from non-fire phenomena. It has lately became clear that fire and non-fire patterns exist as a result of analyzing fire and non-fire testing data. Changing the rate of sensor output per minute and normalized sensor output are used as patterns for analyzing sensors' data in various condition. An odor sensor and a smoke sensor are very effective in discriminating between fire / non-fire phenomena by adopting artificial neural net which has been trained to recognize their patterns.