提出了一种应用可见近红外光谱技术快速测定发动机润滑油动力粘度值的新方法。 对5种不同粘度的润滑油共150个样本进行光谱扫描, 经过光谱预处理后应用偏最小二乘法(PLS)建立了润滑油动力粘度值的预测模型, 并提取出前6个有效主成分作为最小二乘-支持向量机(LS-SVM)预测模型的输入变量, 建立相应的最小二乘-支持向量机(LS-SVM)预测模型, 采用径向基函数(RBF)作为核函数, 超参数γ和RBF核函数参数σ2的最佳组合为γ=27.321 2和σ2=3.229 5。 用125个样本建模, 25个样本验证。 实验结果表明, LS-SVM模型比PLS模型能获得更满意的预测效果。 说明应用光谱技术可以实现发动机润滑油动力粘度值的快速无损检测。

Visible and near infrared (Vis/NIR) spectroscopy was applied for the fast determination of dynamic viscosity of automobile lubricant. One hundred fifty samples from 5 brands were collected for Vis/NIR spectral scanning. Partial least squares (PLS) analysis was applied as calibration method after preprocessing stage as well as a way to extract the first 6 principal components which were used as the input data matrix of least squares-support vector machine (LS-SVM) to develop the LS-SVM models. Radial basis function was used as core function with γ equal to 27.321 2 and σ2 equal to 3.229 5. The calibration set was composed of 125 samples, whereas 25 samples were in the validation set. The results indicated that LS-SVM model achieved the best prediction performance. A new process is proposed in this paper for determination of dynamic viscosity of automobile lubricant.