首页|Prediction of Lubricant Physicochemical Properties Based on Gaussian Copula Data Expansion
Prediction of Lubricant Physicochemical Properties Based on Gaussian Copula Data Expansion
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
万方数据
维普
The composition of base oils affects the performance of lubricants made from them.This paper proposes a hybrid model based on gradient-boosted decision tree(GBDT)to analyze the effect of different ratios of KN4010,PAO40,and PriEco3000 component in a composite base oil system on the performance of lubricants.The study was conducted under small laboratory sample conditions,and a data expansion method using the Gaussian Copula function was proposed to improve the prediction ability of the hybrid model.The study also compared four optimization algorithms,sticky mushroom algorithm(SMA),genetic algorithm(GA),whale optimization algorithm(WOA),and seagull optimization algorithm(SOA),to predict the kinematic viscosity at 40 ℃,kinematic viscosity at 100 ℃,viscosity index,and oxidation induction time performance of the lubricant.The results showed that the Gaussian Copula function data expansion method improved the prediction ability of the hybrid model in the case of small samples.The SOA-GBDT hybrid model had the fastest convergence speed for the samples and the best prediction effect,with determination coefficients(R2)for the four indicators of lubricants reaching 0.98,0.99,0.96 and 0.96,respectively.Thus,this model can significantly reduce the model's prediction error and has good prediction ability.
base oildata augmentationmachine learningperformance predictionseagull algorithm
Feng Xin、Yang Rui、Xie Peiyuan、Xia Yanqiu
展开 >
School of Energy Power and Mechanical Engineering,North China Electric Power University,Beijing 102206,China
北京市自然科学基金Open Project Foundation of State Key Laboratory of Solid Lubrication
万方数据
维普
