Abstract
Distributed neural network based on RBF network is used to establish the model of fire detection in view of complexity of fire process and multiplicity of fire environment. For guaranteeing network Adaptability, the number of subnets neural network and degree of sample belong to subnet are determined by Fuzzy nearest clustering, RBF network is optimized though a improved genetic algorithm by Adaptive Genetic algorithm is improved by Auto-adaptive crossover probability and the mutation probability that determined the adaptation value. The feasibility of adaptability of the model is proved through training and testing in fire detection model