Impact contact behaviors with different velocities, direction and contact states occurred in the process of coal drawing between coal gangue particles and the hydraulic support greatly increase the difficulty of coal gangue recognition in top coal caving. In order to finally realize coal gangue recognition in top coal caving, this paper studied the impact behavior of a single particle coal gangue and the metal plate at any position and the problem of the identification of coal gangue particles. First, the any position impacting test bench and test system is constructed, in which multipoint acceleration information acquisition space is built. Second, 2400 groups of random impact tests between coal gangue particle and the metal plate are conducted. Third, while each impact test was going on, acceleration signals in the seven different positions of the metal plate were collected. Forth, after all the impact tests are completed, signals are processed by segmentation and re-sorted by randomization. Then, the method of signals serial splicing is proposed to process the different position standardized signals obtained in each group of impact test, and coal gangue recognition is proposed based on processed signal data and SVM algorithm. After that, the effect of parameters on coal gangue recognition accuracy of SVM algorithm is studied, and the corresponding parameter settings of SVM are accordingly determined. Finally, coal gangue recognition was conducted by using SVM based on different data samples, the influence law of the training set data set expansion method and the test set data matching relationship, the number of tandem sensors and matching relationship, and the number of training set samples on coal gangue recognition accuracy were studied. The study will serve as the research basis for the application of SVM and signals serial splicing method in coal gangue recognition.

煤石颗粒与水力支架之间在抽煤过程中发生了不同速度，方向和接触状态的冲击接触行为，大大增加了放顶煤对煤recognition石识别的难度。为了最终实现放顶煤中煤石的识别，研究了单颗粒煤石和金属板在任何位置的冲击行为以及识别煤gang石颗粒的问题。首先，构建了影响任意位置的测试台和测试系统，并在其中建立了多点加速度信息获取空间。其次，在煤400石颗粒和金属板之间进行了2400组随机冲击试验。第三，在进行每个冲击测试时，收集了金属板七个不同位置的加速度信号。第四，在完成所有冲击测试之后，信号将通过分段进行处理，并通过随机化进行重新排序。然后，提出了信号串行拼接的方法来处理在每组冲击试验中获得的不同位置的标准化信号，并基于处理后的信号数据和支持向量机算法提出了煤石的识别方法。之后，研究了参数对支持向量机算法对煤石识别精度的影响，并据此确定了支持向量机的相应参数设置。最后，基于不同的数据样本，训练集数据集扩展方法的影响规律和测试集数据匹配关系，通过支持向量机进行煤gang石识别。研究了串联传感器的数量和匹配关系，以及煤set石识别精度的训练集样本数量。该研究将为支持向量机和信号串行拼接方法在煤石识别中的应用提供研究依据。