风机齿轮箱润滑油抗氧化性能预测模型研究  

Research on prediction model of antioxidant performance of wind turbine gearbox lubricating oil

作　　者：底广辉 胡远翔 司明宇 王浩宇 曹俊磊 康举 DI Guanghui;HU Yuanxiang;SI Mingyu;WANG Haoyu;CAO Junlei;KANG Ju(North China Electric Power Research Institute Co.,Ltd.,Beijing 100045,China;School of Mechanical Engineering,Beijing Institute of Petrochemical Technology,Beijing 102617,China;State Grid Jibei Zhangjiakou Wind and Solar Energy Storage and Transportation New Energy Co.,Ltd.,Zhangjiakou 075000,China;State Key Laboratory of Tribology in Advanced Equipment,Tsinghua University,Beijing 100084,China)

机构地区：[1]华北电力科学研究院有限责任公司,北京100045 [2]北京石油化工学院机械工程学院,北京102617 [3]国网冀北张家口风光储输新能源有限公司,河北张家口075000 [4]清华大学高端装备界面科学与技术全国重点实验室,北京100084

出　　处：《应用化工》2024年第11期2517-2523,共7页Applied Chemical Industry

基　　金：国家自然科学基金(52175286);华北电科院自有资金项目(kjz2023019);清华大学高端装备界面科学与技术全国重点实验室开放基金(SKLTKF20B16)。

摘　　要：旨在建立基于红外光谱法快速测定风机齿轮箱润滑油(齿轮油)抗氧化性能的方法。基于齿轮油的红外光谱数据,依次进行样本集划分、数据预处理、特征波长选择和机器学习等数据处理,最终采用多种评价指标对组合模型的性能进行综合评估。结果表明,采用标准正态变量变换(SNV)预处理后的光谱数据所建立的偏最小二乘回归模型性能最佳;两种特征波长提取方法中,主成分分析(PCA)降维效果优于连续投影算法(SPA);三种机器学习中,BP神经网络预测效果最佳。最终得出采用SNV+PCA+BP模型预测效果最优,可以更好地快速预测风机齿轮油的抗氧化性能。This work aims to establish a method for rapidly determining the antioxidant properties of lubricating oil(gear oil)in wind turbine gearboxes using infrared spectroscopy.Based on the infrared spectral data of wind turbine gear oil,a series of data processing steps including sample set partitioning,data preprocessing,characteristic wavelength extraction and machine learning were sequentially performed.Finally,a variety of evaluation indexes were used to comprehensively evaluate the performance of the combined model.The results indicate that the partial least squares regression model established using spectral data preprocessed with standard normal variate(SNV)transformation performs the best.Among the two feature wavelength extraction methods,principal component analysis(PCA)demonstrates superior dimensionality reduction compared to the successive projections algorithm(SPA).Among the three kinds of machine learning,BP neural network has the best prediction effect.The final result indicates that the SNV+PCA+BP model has the best prediction effect,which can better and quickly predict the oxidation resistance of wind turbine gear oil.

关 键 词：红外光谱 机器学习 风机齿轮油 抗氧化性能 模型预测 

分 类 号：TQ646[化学工程—精细化工]