• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文期刊>Mechanical systems and signal processing >Diagnosis of combined faults in Rotary Machinery by Non-Naive Bayesian approach
【24h】

Diagnosis of combined faults in Rotary Machinery by Non-Naive Bayesian approach

机译：基于非朴素贝叶斯方法的旋转机械组合故障诊断

获取原文
获取原文并翻译 | 示例
           
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

When combined faults happen in different parts of the rotating machines, their features are profoundly dependent. Experts are completely familiar with individuals faults characteristics and enough data are available from single faults but the problem arises, when the faults combined and the separation of characteristics becomes complex. Therefore, the experts cannot declare exact information about the symptoms of combined fault and its quality. In this paper to overcome this drawback, a novel method is proposed. The core idea of the method is about declaring combined fault without using combined fault features as training data set and just individual fault features are applied in training step. For this purpose, after data acquisition and resampling the obtained vibration signals, Empirical Mode Decomposition (EMD) is utilized to decompose multi component signals to Intrinsic Mode Functions (IMFs). With the use of correlation coefficient, proper IMFs for feature extraction are selected. In feature extraction step, Shannon energy entropy of IMFs was extracted as well as statistical features. It is obvious that most of extracted features are strongly 'dependent. To consider this matter, Non-Naive Bayesian Classifier (NNBC) is appointed, which release the fundamental assumption of Naive Bayesian, i.e., the independence among features. To demonstrate the superiority of NNBC, other counterpart methods, include Normal Naive Bayesian classifier, Kernel Naive Bayesian classifier and Back Propagation Neural Networks were applied and the classification results are compared. An experimental vibration signals, collected from automobile gearbox, were used to verify the effectiveness of the proposed method. During the classification process, only the features, related individually to healthy state, bearing failure and gear failures, were assigned for training the classifier. But, combined fault features (combined gear and bearing failures) were examined as test data. The achieved probabilities for the test data show that the combined fault can be identified with high success rate.
机译：当旋转机械的不同部分发生组合故障时，它们的功能将具有深远的依赖性。专家完全熟悉单个故障的特征，并且单个故障可以获取足够的数据，但是当故障合并且特征分离变得复杂时，就会出现问题。因此，专家们无法声明有关合并故障的症状及其质量的确切信息。为了克服这个缺点，提出了一种新颖的方法。该方法的核心思想是在不使用组合故障特征作为训练数据集的情况下声明组合故障，而在训练步骤中仅应用单个故障特征。为此，在数据采集并重新采样所获得的振动信号之后，利用经验模式分解（EMD）将多分量信号分解为固有模式函数（IMF）。利用相关系数，选择适当的IMF用于特征提取。在特征提取步骤中，提取了IMF的Shannon能量熵以及统计特征。显然，大多数提取的特征都是高度依赖的。为了考虑此问题，任命了非朴素贝叶斯分类器（NNBC），该分类器释放了朴素贝叶斯的基本假设，即特征之间的独立性。为了证明NNBC的优越性，还使用了其他相应的方法，包括Normal Naive Bayesian分类器，Kernel Naive Bayesian分类器和反向传播神经网络，并对分类结果进行了比较。从汽车变速箱收集的实验振动信号被用来验证该方法的有效性。在分类过程中，仅分配了与健康状态，轴承故障和齿轮故障分别相关的功能来训练分类器。但是，将组合故障特征（齿轮和轴承组合故障）作为测试数据进行了检查。测试数据的概率表明，组合故障的识别率很高。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号