火灾气体辨识的人工神经网络方法研究  

Research on the fire gases identification method based on artificial neural network

作　　者：赵建华[1] 喻益超[1] 李庄[1] 

机构地区：[1]中国科学技术大学火灾科学国家重点实验室,安徽合肥230026

出　　处：《消防科学与技术》2006年第5期675-678,共4页Fire Science and Technology

摘　　要：为了解决火灾气体探测时,能有效地辨识火灾标示性气体的问题,提出一种人工神经网络的辨识方法。通过设计三层反向传播(BP)人工神经网络模型,使用自适应学习速率的BP算法,对所设计的火灾气体辨识BP网络进行训练,经过410次训练后,达到目标误差1×10-4,而且验证实验中的最大输出误差仅为0.0907。表明:火灾气体辨识的人工神经网络方法能有效地识别“火灾”与“非火灾”状态。For the purpose of overcoming the problem of identifying the fire representative gases when detecting the fire gases, an identification method of fire gases based on artificial neural network is discussed. After designing 3 layers back propagation artificial neural network model, we trained the BP artificial neural network by self-adjusting learning velocity BP arithmetic. Four hundred and ten times training resulted in output error less than 1 × 10-1. The verification experiments showed that the largest output error is less than 0. 090 7 and proved that this method is effective in identifying the fact between fire and non fire.

关 键 词：火灾气体探测 人工神经网络 BP算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] X913[自动化与计算机技术—计算机科学与技术]