优化基于近红外光谱的联合间隔偏最小二乘法建模检测芝麻油掺伪含量

Detection of adulteration content in sesame oil by optimizing the model established by synergy interval partial least squares based on near infrared spectroscopy

应用近红外光谱(NIR)分析技术建立测定芝麻油中大豆油含量的定量分析模型。基于32个含量梯度共384个掺伪芝麻油样品的近红外光谱,首先采用标准正态变量变换(SNV)对光谱进行预处理,再采用无信息变量消除法(UVE)初步筛选波长变量,然后结合联合间隔偏最小二乘法(SiPLS)和带极值扰动的简化粒子群优化算法(tsPSO)建立芝麻油中大豆油掺伪含量预测模型,经特征波段选取后建立的模型变量减少,波长变量由451个减少到219个,训练集和测试集相关系数分别为0.9998和0.9919,均方根误差分别为4.39E-2和3.99E-2。结果表明,该方法能够作为芝麻油中大豆油掺伪含量的快速检测方法。此外,该方法也可应用到芝麻油中掺入其他低价值油的掺伪含量检测中。

A quantitative analysis model for determining the adulteration content of soybean oil in sesame oil was established by near infrared spectroscopy technique.Based on the near infrared spectroscopy of a total of 384 adulterated sesame oil samples from 32 content gradients,the spectrum was pre-processed by the standard normal variate transformation(SNV),then the wavelength variable was initially filtrated by the uninformative variables elimination(UVE)method firstly,and then the model for predicting the adulteration content of soybean oil in sesame oil was established by combining the synergy interval partial least squares(SiPLS)with extremum disturbed simple particle swarm optimization(tsPSO).After wavelengths extraction,the number of wavelengths was reduced from 451 to 219.The correlation coefficients of the train set and the test set were 0.9998 and 0.9919 respectively.The root-mean-square errors were 4.39E-2 and 3.99E-2 respectively.The experimental results showed that the method could be used as a rapid method to detect the adulteration content of soybean oil in sesame oil.In addition,the method could also be applied to the detection of adulteration content of other low-value oil in sesame oil.