摘要
利用贝叶斯网络处理不确定性问题能力强和粗糙集约简能够去除冗余性特征的优势,提出了一种基于贝叶斯网络和粗糙集的信息融合方法。该方法提取齿轮泵振动信号的幅域量纲参数作为来自不同传感器的多源信息,改进了特征属性约简方法,设计了贝叶斯网络分类器,构建了多故障贝叶斯网络对特征进行融合,通过最大后验概率准则识别故障类型。两次融合结果对比分析表明,特征属性约简后诊断正确率明显提高,验证了该方法的有效性和实用性。
Making use of advantages that Bayesian network has strong capability of processing uncertain problem and rough set re- duction can eliminate redundant features, an information fusion approach is presented based on Bayesian network and rough set. The vibration signal amplitude domain dimension parameters of gear pump are extracted as multi-source information from different sensors. Then the feature attribute reduction method is improved and the Bayesian network classifier is designed. On the basis of the above, multi-fauh Bayesian network is built up to fuse the features and recognize the fault pattern through the maximum poste- rior probability rule. The contrast analysis of twice fusion results shows that the diagnosis exactitude rate evidently increases after feature attribute reduction. Finally, it proves the validity and practicability of this new information fusion approach.
出处
《现代制造工程》
CSCD
北大核心
2013年第1期125-129,共5页
Modern Manufacturing Engineering
关键词
信息融合
贝叶斯网络
粗糙集
故障诊断
information fusion
Bayesian network
rough set
fault diagnosis
