采煤机作为煤炭开采的主要装置之一，其健康状态受工作环境恶劣、操作空间狭窄等因素影响而难以准确监测，且极易受到煤岩的冲击而发生故障，直接影响采煤机工作效率。此外，由于采煤机特殊的工作环境，使采集的振动数据极易受到各种因素的干扰而变的难以使用，影响其监测可靠性和智能化水平。为准确监测采煤机健康状态，以采煤机正常状态下的电流、温度、流量等容易获取的数据为基础，综合考虑数据集的全局和局部特征以避免其结构信息的丢失。利用主成分分析PCA (Principal Component Analysis)和局部保持投影LPP (Locality Preserving Projections)构建的目标函数，结合互信息、核函数、区间内积估计和重构贡献的方法建立了一种基于简化区间核全局−局部特征融合SIKGLFF(Simplified interval kernel global–local feature fusion)的智能故障诊断方法，用于对表征采煤机状态的非线性不确定数据进行特征提取。并使用山西斜沟煤矿采煤机实际运行数据模拟故障和沙曲二号煤矿实际故障数据对所提方法的性能进行评估实验。结果表明，与中点−半径核PCA、核局部保持投影和区间核全局−局部特征融合算法相比，所提方法对采煤机的单变量模拟故障、多变量的截齿损耗和水路堵塞故障具有良好的监测效果，其故障监测准确率分别达到了99.90%、99.40%和98.70%，计算时间分别只有0.324、0.367和0.345 s，而且可以准确识别产生故障的相关变量，为采煤机故障位置的确定提供理论依据，也为工作人员维护性决策的准确实施指明了方向。

As one of the main equipment in coal mining, the health status of shearer is difficult to be accurately monitored due to some factors such as harsh working environment and narrow operating space, and it is highly susceptible to the impact of coal and rock, which can cause faults and directly affect the working efficiency of shearer. In addition, due to the special working environment of shearer, the vibration data collected is very easy to be interfered by various factors and difficult to be used, which affects the reliability and intelligence level of its monitoring. To accurately monitor the health state of shearer, based on easily available data such as current, temperature and flow under normal state of shearer, the global and local characteristics of the dataset are comprehensively considered to avoid the loss of structural information. The objective function constructed by principal component analysis and local retention projection is used, and an intelligent fault diagnosis method based on a simplified interval kernel global-local feature fusion is established by combining mutual information, kernel function, interval product estimation and reconstruction contribution methods, which is used for the feature extraction of nonlinear uncertain data characterizing the state of coal mining machines. An experimental evaluation is carried out to assess the performance of the proposed method by using the simulated faults of the actual operational data of shearers in the Shanxi Xiegou Coal Mine and the actual fault data of the Shaqu No. 2 Coal mine. The results show that compared with the midpoint-radius kernel PCA, the kernel local preservation projection and interval kernel global-local feature fusion algorithms, the proposed method has a good monitoring effect on the single variable simulation fault, the multi-variable pick loss and the waterway blockage fault of shearers. Its fault monitoring accuracy reaches 99.90%, 99.40% and 98.70%, respectively, and its calculation time is only 0.324, 0.367 and 0.345 s respectively. Moreover, it can accurately identify the relevant variables that cause faults, which provides a theoretical basis for determining the fault location of shearer and also points out the direction for the accurate implementation of maintenance decisions.