基于极限学习机和优化算法的润滑油添加剂种类识别与含量预测

Identification and Content Prediction of Lubricating Oil Additives Based on Extreme Learning Machine

为了快速识别润滑油中添加剂种类和含量,将添加剂硫化异丁烯(T321)、烷基二苯胺(T534)、硫代磷酸胺盐(T307)以不同配比混合在基础油中,使用极限学习机(ELM)对油样的红外光谱数据构建模型进行训练测试,并采用贪心算法、遗传算法(GA)对输入波段优化,筛选出最优波段区间组合以剔除相关性过高的波段从而提高运算效率.测试结果表明:ELM模型可对润滑油添加剂进行有效的种类识别和含量预测,相比于传统理化检测方法是一种经济快速的新型润滑油添加剂检测手段;且经GA波段筛选优化后模型输出结果更具优势,对三种添加剂的种类识别准确率均达到100%、含量预测决定系数(R^2)分别提升了43.8%、39.0%和24.4%.

In order to quickly identify the type and content of additives in the lubricating oil, the lubricating oil additive Vulcanized isobutylene(T321), Alkylated diphenylamines(T534), and Ammonium thiophsphonate(T307) were mixed in the base oil at different proportions. Using an extreme learning machine trains and tests the infrared spectral data construction model of the mixed additive, the greedy algorithm and genetic algorithm were used to optimize the input band, while the optimization results were compared. The optimal band interval combination was selected to eliminate the high excessive correlation band and improve the computational efficiency. The test results showed that the ELM model can effectively identify the type and predict content of lubricant additives. Compared with the traditional physical and chemical detection methods, it was an economical and rapid new lubricant additive detection method. And after the genetic algorithm band filtering optimization, the model output was better. The accuracy of identification of three additives reached 100% after band screening, and the content prediction determinant coefficient(R2) increased by43.8%, 39.0% and 24.4%, respectively.