PSO-GRNN算法辅助拉曼光谱解译牡丹籽油的复杂掺假行为  

Interpretation of complex adulteration behavior of peony seed oil by Raman spectroscopy assisted by PSO-GRNN algorithm

作　　者：邵换峥 孙继红 刘奇付 张晓娟 刘世明 张海华 SHAO Huanzheng;SUN Jihong;LIU Qifu;ZHANG Xiaojuan;LIU Shiming;ZHANG Haihua(Department of Information Technology,Luohe VocationalCollege of Food,Luohe 461002,China;School of Information and Engineering,Zhengzhou University,Zhengzhou 450001,China;College of Information and Management Science,Henan Agricultural University,Zhengzhou 450002,China)

机构地区：[1]漯河食品职业学院,信息工程系,河南漯河462000 [2]郑州大学,信息工程学院,河南郑州450001 [3]河南农业大学,信息与管理科学学院,河南郑州450003

出　　处：《光散射学报》2024年第4期427-435,共9页The Journal of Light Scattering

基　　金：国家自然科学基金(61903340);河南省教育系统党建创新项目(2023-DJXM-200)。

摘　　要：目前针对牡丹籽油掺假量化的研究工作仍集中于单一普通植物油的掺假行为,因此有必要考虑更加复杂的掺假造假情况。本文采用便携式近红外拉曼光谱技术结合粒子群优化广义回归神经网络算法(PSO-GRNN)助力解译牡丹籽油复杂掺假行为。首先,采用廉价葵花籽油和玉米油的混合油品作为掺假对象,并制备涵盖由低到高的牡丹籽油掺假浓度,且浓度梯度变化相对均匀。其次,使用便携式近红外拉曼光谱仪分别采集所有油品样本的拉曼光谱信号,并基于谱学分析对拉曼光谱进行人工降维和权重修正。最后,建立基于PSO-GRNN算法的定量分析模型。结果表明基于便携式近红外拉曼光谱技术结合PSO-GRNN算法可以实现对牡丹籽油复杂掺假行为的解译,该方案不仅可以实现对牡丹籽油浓度的预测,还可以有效评估多种掺假廉价植物油的浓度,其中测试集R^(2)>0.94,RMSE<0.036。本文深入研究了牡丹籽油可能存在的复杂掺假行为并提出了相应的解决方案,该方法对牡丹籽油市场监管和品质检测具有一定的借鉴意义。The research on the quantification of peony seed oil adulteration is still focused on the adulteration of a single cheap vegetable oil,so it is necessary to consider more complex adulteration.This paper uses portable near-infrared Raman spectroscopy technology combined with particle swarm optimization generalized regression neural network algorithm(PSO-GRNN)to help interpret the complex adulteration behavior of peony seed oil.Firstly,the mixed oil of cheap sunflower seed oil and corn oil was used as the adulteration object.The adulteration concentration of peony seed oil was prepared from low to high,and the concentration gradient was relatively uniform.Secondly,a portable near-infrared Raman spectrometer was used to collect the Raman spectrum signals of all oil samples.The Raman spectra were manually reduced and weighted based on spectral analysis.Finally,a quantitative analysis model based on PSO-GRNN algorithm was established.The results show that the complex adulteration behavior of peony seed oil can be interpreted based on portable near-infrared Raman spectroscopy technology combined with the PSO-GRNN algorithm.This scheme can predict the concentration of peony seed oil and effectively evaluate the concentration of various cheap vegetable oils.The test set of the model R^(2)>0.94,RMSE<0.036.This paper deeply studied the possible complex adulteration of peony seed oil and proposed the corresponding solutions.This method is significant for the market supervision and quality detection of peony seed oil.

关 键 词：便携式近红外拉曼光谱 PSO-GRNN 牡丹籽油 掺假 

分 类 号：O657.3[理学—分析化学] O433[理学—化学]