随着机械科学技术的不断发展，对滚动轴承的要求也越来越高，需要其承受更大的载荷和更高的转速，因此急需自主研发高端滚动轴承，突破轴承研发中的核心技术瓶颈，为中国高端装备走出国门、走向世界提供有力的支撑。圆锥滚子轴承由于其特殊结构能够同时承受轴向负荷和径向负荷，被广泛应用在直升机齿轮箱、高速铁路列车轴箱等关键部位。目前，针对圆锥滚子轴承复杂实际服役工况下混合润滑状态的研究还有待加强，圆锥滚子的母线轮廓还有待进一步优化，因此本研究针对圆锥滚子轴承混合润滑状态展开了优化及预测研究。首先，本研究建立了圆锥滚子轴承拟静力学模型与高副接触混合润滑分析模型，并通过与文献结果及实验结果进行对比，验证了所建立的分析计算模型的准确性。在此基础上，针对滚子-滚道摩擦副，为减小圆锥滚子端部出现的粗糙峰接触压力集中的现象，重点分析了不同圆锥滚子修形轮廓对单一滚子和轴承整体混合润滑状态，结合经典Archard磨损理论，提出了面向圆锥滚子的非对称数值磨合修形方法，同时探讨了不同径向载荷、转速、粗糙度等因素对数值磨合修形轮廓以及粗糙峰接触压力分布的影响，数值计算结果表明该方法得到的轮廓能够有效降低滚子端部粗糙峰接触压力，为圆锥滚子母线轮廓优化设计提供新的思路。接着，针对滚子端部-轴承挡边摩擦副，通过高副接触滑动摩擦实验以及数值计算，探究了磨合后摩擦副轮廓对高副接触滑动摩擦的影响，阐明了实现高副接触低滑动摩擦的机理和必要条件：接触区域从接触点到类冠状磨斑的转变、磨合后形成了纳米级粗糙度、摩擦副相对运动处在合适的工况条件，同时分析了接触半宽、轮廓倒圆角半径等因素等影响，为滚子端部-挡边摩擦副轮廓优化提供指导。最后，应用神经网络方法，对圆锥滚子混合润滑状态和数值磨合修形轮廓进行了预测分析，针对膜厚、压力等润滑状态参数，对比了神经网络模型预测结果和数值计算结果的相对误差，总结了误差产生的原因以及减小误差的有效途径，比较了数值计算算例平均耗时与神经网络模型训练及预测耗时，实现了对任意工况下数值磨合修形轮廓的预测，证明了神经网络方法在圆锥滚子混合润滑状态和数值磨合修形轮廓方面的可行性，显示了神经网络在轴承中更多的应用前景及优势。
With the continuous development of mechanical science and technology, the demand for rolling bearings is also increasing, requiring them to withstand larger loads and higher speeds. Therefore, there is an urgent need to independently develop high-end rolling bearings, break through the core technological bottleneck in bearing research and development, and provide strong support for China‘s high-end equipment to go abroad and go global. Tapered roller bearings are widely used in key parts such as helicopter gearboxes and high-speed railway train axle boxes due to their special structure that can withstand axial load and radial load simultaneously. At present, research on the state of mixed lubrication of tapered roller bearings under practical complicated operation conditions still needs to be strengthened, and further research is needed to optimize the profile and shape of tapered rollers. Therefore, this study focuses on optimizing and predicting the mixed lubrication state of tapered roller bearings.Firstly, this study established a quasi-static model for tapered roller bearings and a high pair contact mixed lubrication analysis model, and verified the accuracy of the established model through comparison with other calculation results in literature and experimental measurement results. Building upon this, for the roller-raceway friction pair, to mitigate the phenomenon of concentrated contact pressure spikes at the ends of tapered rollers, the study focused on analyzing the effect of different tapered roller profile modifications on the mixed lubrication state for both individual rollers and the overall bearing. Combining the classical Archard wear theory, an asymmetrical numerical conformal modification method for tapered rollers was proposed. Simultaneously, the study investigated the influence of various factors such as radial load, speed, and roughness on the numerical conformal modification profile and the distribution of contact pressure spikes. The profiles obtained through this method effectively reduce the contact pressure spikes at the ends of the rollers, providing new insights for the design of the base profiles of tapered roller bearings.Furthermore, concerning the end-of-roller to bearing rim friction pair, the study conducted high-load contact sliding friction experiments and simulation calculations to investigate the influence of the post-break-in friction pair profile on high-load contact sliding friction. The study clarified the mechanisms and necessary conditions for achieving high-load contact with low sliding friction: the transition of the contact area from the contact point to a crown-shaped wear mark, the formation of nano-scale roughness after running-in, and the appropriate working conditions for relative motion of the friction pair. Factors such as contact half-width and contour fillet radius were also analyzed. This analysis provides guidance for optimizing the structure of the end-of-roller to rim friction pair.Finally, by using neural network methods, a predictive analysis of the mixed lubrication state and numerical conformal modification profiles for tapered rollers was conducted. Regarding lubrication state parameters such as film thickness and pressure, the relative errors between the predictions of the neural network model and the simulation results were compared. The study summarized the reasons for the errors and effective ways to reduce them. The average computation time of numerical calculations was compared with the training and prediction time of the neural network model. The study achieved the prediction of numerical conformal modification profiles under arbitrary operating conditions, demonstrating the feasibility of neural network methods in predicting the mixed lubrication state and numerical conformal modification profiles for tapered rollers. Based on the research findings, the study elucidated the broader application prospects and advantages of neural network in bearing design.
