本文提出将串联型模糊神经网络应用于火灾探测系统。首先将输入信号模糊化后送入神经网络进行处理，其模糊系统用指派法定义隶属度函数，神经网络采用三层前馈ＢＰ网络结构，用ＢＰ算法进行网络参数的训练。其后由模糊逻辑判决火灾或非火灾。实验结果表明该网络能够准确探测各种标准试验火并有很强的抗干扰能力。 This paper proposes the application of series fuzzy neural network to fire detection systems. First, the input signal is blurred and sent to the neural network for processing. The fuzzy system uses membership method to define the membership function. The neural network adopts a three-layer feed-forward BP network structure and uses BP algorithm to train the network parameters. Afterwards, fire or non-fire was decided by fuzzy logic. Experimental results show that the network can accurately detect a variety of standard test fires and have a strong anti-jamming capability.