基于孪生网络和自编码网络结合的时间分辨拉曼光谱油料成分分析  

Time-Resolved Raman Spectroscopic Analysis of Fuel Components Based on the Combination of Twin Network and Autoencoder Network

作　　者：豆西博 李琮举 王昊 Dou Xibo;Li Congju;Wang Hao(Key&Core Technology Innovation Institute of The Greater Bay Area,Guangzhou 510535,China;Zhongke Atomically Precision Manufacturing Technology Co.,Ltd,Dongguan 523808)

机构地区：[1]广东粤港澳大湾区硬科技创新研究院,广东广州510535 [2]东莞市中科原子精密制造科技有限公司,广东东莞523808

出　　处：《实验与分析》2024年第4期34-38,共5页LABOR PRAXIS

基　　金：广东省重点领域研发计划项目(2019B090917012)。

摘　　要：基于拉曼光谱的指纹图谱对燃料油中化合物含量值分析,提出了一种结合孪生网络(SN)和自编码网络(AE)的燃料油化合物检测模型,用于定量预测训练阶段未包含的油料化合物含量。该检测模型利用SN权重共享的特点对不同型号燃油的时间分辨拉曼光谱,通过对待测油料的特征进行分类,以完成相似油料类型数据选取。In the article,a fuel oil compound detection model that combines the twin network(SN)and the autoencoder network(AE)is put forward based on the fingerprint of the Raman spectrum.This model is designed to quantitatively predict the content of oil compounds which are not included in the training phase.The detection model takes advantage of the feature of weight sharing in the SN to analyze the time-resolved Raman spectroscopy data of various types of fuel oils.the selection of data belonging to similar oil types is accomplished.

关 键 词：时间分辨拉曼光谱 孪生网络 自编码网络 油料成分 

分 类 号：TQ517.4[化学工程]