摘要
为降低火灾自动报警系统的误报、漏报率,基于BP神经网络算法,用LabVIEW虚拟仪器开发了一套智能火灾识别模型。在火灾探测区域内合理布置若干感温探测器,在火灾识别模型中,将探测到的温度场参数作为BP神经网络的输入,火灾发生与否作为输出,并对影响BP神经网络的各项参数和该模型的运行结果进行测试研究。仿真试验结果表明:选取42组训练样本,当网络训练到4 000次左右时,最大相对误差值达到目标值0.1,其中大部分相对误差值达到0.05以下,网络的实际输出值非常逼近样本的理想输出值;实际火灾试验表明:该火灾识别模型能够探测火灾的发生。
To reduce the false and leakage alarm rate of automatic fire alarm system,an intelligent fire identification model based on BP neural network algorithm was developed by using LabVIEW virtual instrument.In this model,several thermal detectors were arranged in the fire scene appropriately;the fire field parameters detected by the thermal detectors were used as the input values of BP neural network,and the situation of fire occurrence was used as the output value.Meanwhile,the effect of every parameter on the BP neural network and on the running results of the model was tested.Simulation experiment on 42 training samples shows that,when the network training times reach about 4 000,the maximum relative error value gets to the target value 0.1;most of the relative errors are below 0.05,and the actual output value of the network approach the ideal output value of the samples.Actual fire experimental results show that the fire identification model can detect the occurrence of fire.
出处
《中国安全科学学报》
CAS
CSCD
北大核心
2011年第11期49-55,共7页
China Safety Science Journal
基金
天津理工大学教学基金资助(YB10-05)
