Abstract:

Aiming at the problems of false alarm and missing alarm in traditional fire prediction methods, a fire prediction method based on adaptive integrated neural network is proposed. Firstly, the rate detection algorithm is used in the information layer to input the singular data detected by different types of sensors into the network model. Secondly, in the feature layer, the long- term memory network (LSTM) and radial basis function feed forward neural network (RBF-BPNN) are used to build an integrated network to learn the fire characteristics under different input parameters. Finally, the fuzzy logic control system is designed at the decision-making level to infer and output the fire alarm level. Experimental results show that this method has higher prediction accuracy.  

Key words:

adaptive ensemble neural network, rate detection algorithm, fuzzy logic control system, fire prediction