润滑油品质近红外及拉曼光谱检测方法.pdfVIP

摘要 对掺加量校正模型的输入变量优化。研究结果为，润滑油掺假量模型的相关系数均大于 0.972， RMSEP 均小于 0.0277，能很好地对验证集进行预测。采用 iPLS 波段筛选方法 -1 对原始波段进行特征变量筛选，选出了 1148-1484cm 波段做为拉曼光谱定量检测润滑 油掺假的特征谱段，利用 GA 选出了 185 个波数点。选用昆仑天歌掺假美孚润滑油作为 特征波段筛选验证并取得较好的结果。 （6 ）研究了润滑油粘度拉曼光谱定量检测方法。利用iPLS 和 GA 算法对润滑油原 始光谱进行特征变量筛选。研究结果为，采用 iPLS 筛选的特征波段建模效果要优于 GA 遗传算法筛选得到的结果，模型相关系数 Rp 为 0.963，预测集均方根误差 RMSEP 为 10.1078。能够很好的对润滑油的运动粘度进行检测。 本论文研究表明采用近红外和拉曼光谱技术结合 PLS 建模算法对润滑油的品质（含 水量、粘度）、掺假以及品牌鉴别是可行的。采用一定的光谱波段筛选方法可以大大减 少建模的输入变量，减少计算过程，提高模型的稳健性和准确度。 关键词：光谱技术；润滑油；粘度；水分；品牌；iPLS ；GA II Abstract THE LUBRICANT QUALITY NEAR-INFRARED AND RAMAN SPECTROSCOPY TESTING METHODS ABSTRACT Lubricant oil is well known as the ‘blood’ of automobile engine, it relates to the health status of the engine. The suitable lubricants not only be able to guarantee the normal operating state and can also extend its life and performance. Nowadays, there are many brands of lubricants with different qualities. It is difficulty to distinguish between good and inferior from the outside. The quality of lubricant is a reference index of whether to replace the oil on the care and maintenance. The traditional way of test the brand and quality of the lubricant oil is rely on experience or chemical instruments to chemical composition analysis, this process is not only cost time but also may cause pollution. So it is necessary to study a fast, non-destructive, economy method to detecting the quality of the lubricant oil. In this paper, near infrared spectroscopy and Raman spectroscopy are applied to study the feasibility of lubricant brand and the quality indicators (viscosity, water content), ana