In order to reduce the leak-check and error rate of the fire alarm system

the nonlinear reflecting capability of neural network is applied to intelligently process the data synchronously detected by multi-sensors（the temperature sensor

the smog sensor and the CO sensor）.The result of simulation shows that the different kinds of signals related to fire disaster can be accurately recognized by the fire alarm system with multi-sensors based on neural network

and the false-alarm can be decreased.The suitability of this system to environmental variation and anti-interference capability can be obviously improved.