摘要
根据催化传感器在不同的电场条件下具有不同气体检测灵敏度的特点,介绍了采用单一的热催化传感器在不同的电场强度下,通过分布式多子网神经网络对含未知气体的可燃混合气体进行分析的新方法。应用分布式神经网络,通过训练建立了信号识别的模型,并以3种混合气体为对象进行实验,结果证明了分析方法的可行性。实验表明:该网络在泛化能力与学习速度等均优于BP和RBF网络,其多子网、自动分解任务的特点尤其适用于复杂样本的学习,具有很好的应用前景。
Based on the feature of the sensitivity detection of catalytic sensor to inflammble different gas, single catalytic sensor is used to analyze mixed-inflammable gases with unknown components by distributed multi-subnet neural networks in different electric field. Distributed neural networks are used to set model of signal recognition by training. Three mixed gases are adopted. The result proves that the method is feasible. And the result shows that it achieves better performance than BP or RBF network on generalization ability and training speed etc. The feature of multi-subnet and dividing a complex task automatically is especially suitble for complex task learning. And it has good application prospect.
出处
《传感器与微系统》
CSCD
北大核心
2009年第11期16-18,共3页
Transducer and Microsystem Technologies
基金
国家自然科学基金资助项目(50374067)
关键词
气体分析
催化传感器
分布式多子网神经网络
gas analysis
catalytic sensor
distributed multi-subnet neural networks
