摘要
基于光纤—电容液滴分析方法的原理,研究了盐度的检测。由20种不同盐度溶液的实验结果得出:随着盐度的增加,光纤信号值和液滴指纹图曲线下的面积不断增大。利用1—12—1三层BP神经网络结构,分别以光纤信号的平均值和液滴指纹图曲线下的面积为输入向量,进行网络训练,并利用训练好的网络分别进行测试。比较测试结果得出:利用光纤信号的平均值作为BP网络的输入向量进行盐度检测效果较好,最大检测误差为0.14%。
Based on the principle of fiber-capacitance drop analysis method, salinity detection is studied. The experimental results of 20 kinds of solution with different salinity show that the optical fiber signal value and area under liquid droplets fingerprints diagram curves increases with increase of salinity. Using 1-12-1 three-layer BP neural network structure, and average value of the optical fiber signal and the area under the curve of the liquid droplets fingerprints as input vectors respectively, and BP neural network is trained and uses the trained network to test respectively. It is derived by comparison of test results, salinity detection effect using mean value of optical signal as input vector of the BP networks is better than the others, and the maximum detecting error is 0.14 %.
出处
《传感器与微系统》
CSCD
北大核心
2013年第7期123-125,135,共4页
Transducer and Microsystem Technologies
基金
中央高校基本科研业务费专项资金资助(DL12BB27)
关键词
光纤—电容液滴分析
盐度
液滴指纹图
BP神经网络
fiber-capacitance liquid droplet analysis
salinity
liquid droplets fingerprints figure
BP neural net- work
