应用近红外光谱技术结合支持向量机回归(support vector machine regression, SVR)方法测量食用植物油脂低碳数脂肪酸(C≤14)含量。 使用SupNIR-5700近红外光谱仪采集58个样品的近红外光谱图, 通过偏最小二乘(partial least square, PLS)算法剔除奇异样品。 选择其中具有代表性的52个样品进行主成分分析(principal component analysis, PCA), 选取径向基(radial basis function, RBF)核函数建立支持向量机回归模型, 并对光谱预处理方法和参数寻优方法进行了详细的分析和讨论。 实验表明, 经过粒子群算法(particle swarm optimization, PSO)优化后模型的性能都有所提高, 泛化能力更强, 预测的准确度和稳健性更好; 其中预处理方法2经过PSO优化寻优后的参数C=2.085, γ=22.20时, 预测集和校正集相关系数(correlation coefficient, r)分别达到了0.998 0和0.925 8, 均方根误差(root mean square error, MSE)分别为0.000 4和0.014 3。 研究结果表明, 应用近红外光谱结合PSO-SVR方法进行食用植物油脂低碳数脂肪酸含量快速、 准确的预测是可行的。

The rapid prediction of the low-carbon fatty acids (C≤14) content in grease samples was achieved by a mathematical model established by near infrared spectroscopy combined with support vector machine regression (SVR). In the present project, near-infrared spectrometer SupNIR-5700 was used to collect near-infrared spectra of 58 samples; partial least square (PLS) was applied to remove the strange samples, and principal component analysis (PCA) was conducted on the measurements; radial basis function (RBF) kernel function was selected to establish a regression model supporting vector machine, and then detailed analysis and discussions were conducted concerning their spectral preprocessing and parameters optimization methods. Experimental results showed that by applying particle swarm optimization (PSO) the model demonstrated improved performance, stronger generalization ability, better prediction accuracy and robustness. In the second pretreatment method after PSO, when the optimization parameters are: C=2.085, γ=22.20, the prediction set and calibration set correlation coefficient (r) reached 0.998 0 and 0.925 8, respectively; and root mean square errors (MSE) were 0.000 4 and 0.014 3, respectively. Research results proved that the method based on near infrared spectroscopy and PSO-SVR for accurate and fast prediction of the low-carbon fatty acid content in vegetable oil is feasible.