液压系统是采煤机重要的组成部分，担负着调高和制动的作用。近几年，随着工况自动化水平的不断提高，煤矿生产的机电一体化装备也越来越高，与此同时，采煤机液压系统的故障发生频率也与日俱增。所以，快速有效的对采煤机液压系统进行故障诊断，不但会增进经济效益而且可以预防安全事故。 本文以采煤机为研究对象，对采煤机液压系统故障诊断的研究意义进行了分析，在广泛了解采煤机液压系统国内外研究现状、发展趋势和常用故障诊断方法的基础上，提出了基于数据融合的故障诊断方法。 首先，针对采煤机的基本机械原理分析了采煤机液压系统的液压泵、泵站电机和制动器的故障机理，进而阐述了引起故障的原因和故障引起的结果。 然后，介绍了数据融合的概念，分别提出了基于神经网络和D-S证据理论的故障诊断方法。通过实例发现，前一种方法容易陷入局部最小且初始的权值和阈值选取存在较大的随机性，降低了故障诊断的精确性，所以通过遗传算法对它做进一步优化。 最后，构建了二级采煤机液压系统故障综合诊断模型，利用GA-BP神经网络进行特征级的局部故障诊断，通过对输出结果的相关计算得出测试样本的基本可信度分配函数，应用D-S证据理论对得到的基本可信度分配函数进行融合，得到了比较直观的诊断结果，证实了基于GA-BP和D-S证据理论结合的数据融合故障诊断方法的有效性。

As one of the important component in shearer, hydraulic system has the function of controlling height and brake. Shearer electromechanical integration equipment is higher with the constant improvement of the automation level. At the same time, shearer hydraulic system failure frequency is also increasing. Therefore, the fault diagnosis technology is applied to shearer hydraulic system efficiently and effectively, not only can enhance productivity, but also prevent safety accidents. In this thesis, the significance of shearer hydraulic system fault diagnosis is introduced, fault diagnosis method based on data fusion is proposed on the basis of broad understanding of the shearer hydraulic system research status, development trend and common fault diagnosis method. The basic mechanical principle of shearer is introduced. The failure mechanism is analyzed from hydraulic pumps, motors and brake pump station from shearer hydraulic system. The fault of causes and results are elaborated. Afterwards, the concept of data fusion is introduced simply. The fault diagnosis methods which including neural networks and D-S evidence theory are proposed respectively. Through example, the former is easy to fall into local minimum, more than that, the initial weights and threshold value exist a greater randomness. It can reduce the accuracy of fault diagnosis, so the former needs to be further optimized through the genetic algorithm. Eventually, a secondary fault diagnosis model of shearer hydraulic system is established. The fault is diagnosed locally in characteristic class. The basic probability distribution function of the test sample is calculated through to the calculation related output. It is fused by D-S evidence theory, getting a more intuitive diagnosis. The validity of the method is confirmed.