溜槽是工业应用中的一个关键设备，溜槽堵塞是造成系统事故和降低运行效率的主要因素之一，然而堵塞预防与检测是实际应用中的一个难点。本文在总结研究已有的溜槽堵塞预防与检测方法的基础上，结合机械故障诊断理论，采用溜槽的振动信号识别其运行过程中的状态，在模式识别的应用方面有着重要的学术价值和实际应用意义。 本文将数据挖掘技术应用于溜槽振动模式分析，结合溜槽振动数据的特性，使用特征提取和模式分类方法，对振动信号进行了处理分析，最后开发了振动信号模式分析系统。本文主要内容为： 首先，研究了振动特征提取方法，并结合直方图的相关理论，提出了等宽强度序列和等深强度差序列特征划分方法。并通过输煤系统中的溜槽振动数据对时域和时频域特征提取方法进行了相关实验，实验结果表明，不同模式下的振动数据经过特征提取方法预处理后，具有一定的辨识度。 其次，根据输煤系统的振动特征数据，使用相似度度量、KNN分类方法对实验数据进行了模式分类，首先对源数据进行了特征提取，然后通过数据样本间的相似度，对测试样本进行了分类；使用朴素贝叶斯分类方法，首先对训练样本属性进行分类概率统计，然后将统计结果应用于测试样本对其进行分类。最后使用正确率、错误率、检测率对各分类方法进行了评价。分析结果表明，采用KNN分类方法与等宽强度序列方法的结合使用，进行振动模式分析的性能优于其他两种方法。 最后，在上述研究成果的基础上，应用特征提取和分类方法，开发了溜槽振动信号模式分析系统。系统包括源数据采集模块，源数据特征参数化模块，模式匹配分析模块。实验结果证明系统能有效的识别溜槽振动模式，达到了预期目标。

Chute is a critical equipment in industrial applications. Chute clogging, one of the major factors lead to system incident, is a difficult point to guard against and detect in the practical applications. This thesis summarize the basis of chute clogging researches and prevention methods, use the theory of mechanical failure diagnosis, select the vibrating signal of chute, and identify its processes operation states. This research of topic is of great important academic value and practical significance to the studying of pattern recognition. This thesis applies data mining to the chute vibrational mode analysis. Considering the properties of the data of chute vibration, use the feature extraction and the pattern classification techniques to process the chute vibration data. Finally, develop the analysis system of vibration signal patterns. The main contents of this thesis include the followings: Firstly, considering the theory of histogram, research the extraction methods of vibratory signature, and present the feature classifying processes of equi-height strength sequence and equi-depth strength difference sequence. Experiment with the chute vibration data of the coal handling system by the methods of time and time-frequency domains feature extraction. The results show that, after the pretreatment of feature extraction methods, the vibration data of different models can be identified. Secondly, the similarity measure and KNN classify method are applied to assort the pattern of vibration feature experimental data of the coal handling system. Source data features are extracted, and then the test samples are classified by the similarity between data samples; the naive Bayesian classification method is used, the property of training samples are classified according to the statistic information, then the statistical results are applied to the test samples classification. The classification methods are assessed by accurate rate, error rate and detection rate. The results shows that, it’s more effective to combine the KNN classify method with equi-height strength sequence method than other two methods. Finally, based on the research result stated above, feature extraction and classification methods are applied to present analysis system of chute vibration signal patterns. The system includes the data collection module, the source data feature parameterization module, the pattern matching and analyzing module. The results show that the system can be effectively applied to identify the chute vibration patterns, and achieve the expected goals.