天轮是提升系统中的重要组成部分，一旦其出现故障，不仅会影响煤矿的生产效率，而且会引发安全风险。润滑不良是导致天轮故障的主要原因之一。针对目前煤矿现场普遍的人工润滑方式导致效率低下、难以保证润滑的均匀性和及时性等问题，设计了新型天轮轴承智能润滑系统，并实现其现场应用。从天轮结构出发，考虑天轮轴端滚动轴承和游动轮滑动轴承的不同运行方式，提出了采用自动润滑和人工辅助相结合的方式对天轮轴承进行润滑，通过对润滑系统的实时监测实现单泵多点不同用脂量智能配送功能，同时提供废脂回收功能避免废油对轴承的损伤。研究了基于润滑状态异常识别的智能控制策略，根据对润滑状态的判断结果调整系统的润滑周期和润滑量。考虑到润滑状态识别准确率对系统运行结果的影响，构建了基于局部保留投影算法(Locality Preserving Projections，LPP)与支持向量数据描述(Support Vector Data Description, SVDD)相结合的润滑状态智能识别模型。针对特征降维方法中近邻点数量会严重影响降维效果的问题，提出以样本集的高低维特征集的信息熵差为目标函数，采用粒子群优化算法对LPP算法参数进行优化。采用现场实测数据对该模型进行验证，结果表明所提方法可有效识别天轮轴承异常状态，现场工况下识别准确率达82%。最后，在某煤矿进行了天轮轴承智能润滑系统的现场应用研究，以单泵对4个天轮进行润滑，结果表明润滑系统可实时获取自身的运行参数并实现远程控制，保证润滑效果的持续性和均匀性，满足现场的实际需求。

The sheave is an important part of a mine hoisting system. Once it fails, it will not only affect the production efficiency of coal mine, but also cause some safety risks. Poor lubrication is one of the main causes of sheave failure. In view of the low efficiency and difficulty in ensuring the uniformity and timeliness of lubrication caused by the current manual lubrication method commonly used in coal mines, this paper presented a new intelligent lubrication system for sheave bearings and implemented it on-site application. First of all, starting from sheave structure and considering the different operating modes of the sheave shaft end rolling bearing and the traveling wheel sliding bearing, it was proposed to use a combination of automatic lubrication and manual assistance to lubricate the sheave bearing. Through a real-time monitoring of the lubrication system, it realized the intelligent distribution function of single pump and multiple points with different grease amounts, and also provided the waste grease recovery function to avoid damage to bearings by waste oil. Then, an intelligent control strategy based on the identification of abnormal lubrication status was studied, and the lubrication cycle and lubrication amount of the system were adjusted based on the judgment results of the lubrication status. Subsequently, considering the impact of lubrication status identification accuracy on system operation results, an intelligent identification model of lubrication status based on the combination of Locality Preserving Projections (LPP) and Support Vector Data Description (SVDD) was constructed. In view of the problem that the number of neighboring points in the feature dimensionality reduction method will seriously affect the dimensionality reduction effect, it was proposed to use the information entropy difference between the high and low dimensional feature sets of the sample set as the objective function, and apply the particle swarm optimization algorithm to optimize the parameters of the LPP algorithm. The model was verified using on-site measured data. The results show that the proposed method can effectively identify the abnormal state of the sheave bearing, and the identification accuracy can reach 82% under on-site working conditions. Finally, an on-site application study of the intelligent lubrication system for pulley bearings was conducted in a coal mine. A single pump was used to lubricate four pulleys. The results show that the lubrication system can obtain its own operating parameters in real time and achieve remote control, ensure the continuation and uniformity of the lubrication effect, and meet the actual requirements of the mine site.