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

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(R2)分别提升了43.8%、39.0%和24.4%.

To rapidly identify the types and contents of additives in lubricating oils, additives including sulfurized isobutylene (T321), alkyldiphenylamine (T534), and amine thiophosphate (T307) were mixed into base oil at different ratios. Extreme Learning Machine (ELM) was used to construct and train/test models based on the infrared spectral data of oil samples. Additionally, greedy algorithm and Genetic Algorithm (GA) were employed to optimize the input spectral bands, and the optimal band interval combinations were screened to eliminate highly correlated bands, thereby enhancing computational efficiency. Test results show that the ELM model can effectively identify the types and predict the contents of lubricating oil additives, serving as an economical and rapid novel detection method for lubricating oil additives compared with traditional physicochemical detection methods; moreover, the model output results demonstrate greater advantages after GA-based band screening and optimization: the classification accuracy for the three additives all reached 100%, and the coefficient of determination (R²) for content prediction increased by 43.8%, 39.0%, and 24.4% respectively.

搜集汇总

背景与挑战

背景概述

该数据集聚焦于润滑油添加剂检测，通过极限学习机（ELM）模型结合优化算法（如遗传算法）对红外光谱数据进行分析，实现了添加剂种类识别与含量预测。其特点在于采用波段优化技术提升模型效率，实验结果显示经优化后识别准确率达到100%，且含量预测精度显著提高，为润滑油检测提供了一种经济快速的新方法。