基于灰狼算法优化GRNN的润滑油摩擦磨损性能预测

针对齿轮油极压抗磨添加剂的复配问题，提出基于灰狼算法优化的广义回归神经网络(GWO-GRNN)摩擦学性能参数优化模型. 选用齿轮油常用的硫化异丁烯(T321)、磷酸三甲酚酯(T306)、异辛基酸性硫磷脂十八胺(T308)和二烷基二硫代氨基甲酸钼(MoDTC)这4种材料为添加剂，设计正交试验制备齿轮油并使用MFT-R4000往复摩擦磨损试验机测试其摩擦学性能，分别建立平均摩擦系数和磨损体积性能预测模型并对模型参数进行优化，提高模型预测的准确性，采用留出法和留一交叉验证法评估模型在数据集上的泛化能力，降低模型过拟合的风险. 研究结果表明：在引入灰狼算法(GWO)优化广义回归神经网络(GRNN)的平滑参数σ后，预测模型的决定系数R2得到明显提升，GWO-GRNN平均摩擦系数预测模型的R2达到96%，磨损体积预测模型的R2达到91%；表明该模型能够在小样本情况下较为准确预测出齿轮油极压抗磨添加剂的摩擦学性能，为齿轮油极压抗磨添加剂的复配研究提供了新方法.

Aiming at the compounding problem of extreme pressure anti-wear additives for gear oils, a tribological performance parameter optimization model of generalized regression neural network (GWO-GRNN) optimized by grey wolf optimizer is proposed. Four commonly used additives for gear oils, namely sulfurized isobutylene (T321), tricresyl phosphate (T306), isooctyl acidic thiophosphoric acid octadecylamine (T308) and molybdenum dialkyldithiocarbamate (MoDTC), were selected as the test additives. Orthogonal experiments were designed to prepare gear oil samples, and their tribological properties were tested using an MFT-R4000 reciprocating friction and wear tester. Prediction models for average friction coefficient and wear volume were established respectively, and their model parameters were optimized to improve the prediction accuracy of the models. The hold-out method and leave-one-out cross-validation were adopted to evaluate the generalization ability of the models on the dataset, so as to reduce the risk of model overfitting. The research results show that: after introducing the grey wolf optimizer (GWO) to optimize the smoothing parameter σ of the generalized regression neural network (GRNN), the coefficient of determination R² of the prediction models is significantly improved; the R² of the GWO-GRNN average friction coefficient prediction model reaches 96%, and that of the wear volume prediction model reaches 91%; this indicates that the model can accurately predict the tribological properties of extreme pressure anti-wear additives for gear oils under small sample conditions, providing a new method for the compounding research of extreme pressure anti-wear additives for gear oils.

搜集汇总

背景与挑战

背景概述

该数据集聚焦于润滑油摩擦磨损性能预测，采用灰狼算法优化GRNN模型，通过四种添加剂的正交试验数据建立预测模型，在小样本情况下表现出较高的预测准确性。