基于数据融合与神经网络的火灾探测研究  

A primitive study of fire detection method based on data fusion and BP neural network

作　　者：刘少刚[1] 吕建伟[1] 王士成[1] 

机构地区：[1]哈尔滨工程大学机电工程学院,黑龙江哈尔滨150001

出　　处：《应用科技》2011年第5期9-12,共4页Applied Science and Technology

基　　金：科技部社会公益基金专项资助项目(2005DIB3J138)

摘　　要：海上平台由于其自身空间狭小、管线设备高度集中等特点,要求其火灾探测系统误报率低.目前,单一类型火灾探测器的误报率非常高,经研究发现同时探测多类型火灾因素可大幅度降低误报率.文中根据多传感器数据融合技术将火灾探测器所测数据进行融合,然后应用BP神经网络进行训练仿真,降低了火灾探测器的误报率,满足了海上平台火灾探测系统的要求.The fire detection system of offshore platforms requires a low false alarm rate due to its features such as cramped interior and highly centralized equipment of the pipeline.Now,the false alarm rate of any single type of fire detection is very high.Researches showed that detecting multiple fire factors at the same time could greatly reduce the false alarm rate.Based on applying the integral techniques of multi-sensor data fusion to the fire detection sensors,this thesis has implemented the simulation of BP neural network to reduce the false alarm rate of the fire detection sensor,which meets requirement of the offshore platform＇s fire detection.

关 键 词：多传感器 数据融合 神经网络 火灾预报 

分 类 号：U692[交通运输工程—港口、海岸及近海工程]