基于遗传神经网络的电子鼻在大气环境气体模式识别中的应用

Electronic Nose with Genetic Neural Network for Mixed Gases Pattern Recognition

基于模拟退火遗传神经网络的电子鼻已对二组分、三组分的混合气体模式进行了识别 ,识别精度及学习速度都较 BP神经网络大为提高 [1 ] ,但仍无法满足实用的要求。为了提高遗传神经网络对混合气体越限值的识别精度 ,本文在电子鼻已有的基础上提出分步分档识别法 ,在大范围内保证了识别准确性 ,提高了电子鼻的实用性。本文成功地将其用于四组分混合气体的精确识别。应用本方法的电子鼻既可用于正常环境气氛也可适用于危险气氛环境的气体模式识别。不同的档 ,其学习样本不同 ,识别精度不同。分档识别精度取决于学习样本的最小步长 ,最小步长越小 。

The electronic nose with genetic neural network has recognized the pattern of the mixed gases with two components and three components ,with precision improved and training speed increased compared with BP network.But the recognition precision couldnt reach the specification of applications when the measure range is very large.In order to increase the recognition precision of the over limit values of the mixed gases, and improved method of staged sectioning recognition is proposed in this paper.The improved method can recognize the mixed gases in a large rang accurately, and enhance the practicability of electronic nose.This method has recognized the mixed gases with four components accurately in this paper.The gas pattern recognised is not only in normal range but also in failure condition.The improved method divides the whole range into several sections.The studied specimen and recognition precision are different in different sections.The staged sectioning recognition precision depends upon the minimum interval of the studied specimen.In normal concentration section,the recognition precision doesnt need to be very high, so the interval could be long;while in the critical value section,precision need to be very high,and the interval is short.The smaller the minimum interval of a studied specimen is, the higher the recognition precision is.