基于奇异值分解、模糊聚类和粗糙集理论的旋转机械故障诊断

FAULT DIAGNOSIS OF ROTATING MACHINERY BASED ON SVD,FCM AND RST

对含有重复和冲突对象的离散决策表,提出了一种基于粗糙集的规则获取方法,使得获得的规则能够涵盖所有的对象。对连续条件属性值和离散决策属性值的决策表,基于矩阵的奇异值分解、模糊C均值聚类和粗糙集属性约简技术,提出连续属性最佳离散数目确定方法。在上述方法的基础上,进行旋转机械故障诊断的规则获取,获得的诊断规则具有很好的知识归纳能力和知识泛化能力。利用获得的诊断规则进行旋转机械故障诊断,建立了待诊断对象和诊断规则的弹性匹配模式,使得诊断结论的获取取决于不同的诊断要求。

For decision table (DT) with reduplicate and conflictive objects, a rule acquisition approach based on rough set theory (RST) is presented, which makes the rules acquired cover all the objects in DT. Based on singular value decomposition (SVD) of matrix, fuzzy C-means clustering (FCM) and RST based attribute reduction, an optimal discrete approach of continuous condition attribute values (CCAVs) in DT with CCAVs and discrete decision attribute values is put forward. The above approaches are utilized for rule acquisition in fault diagnosis of rotating machinery, whereby the acquired rules have not only good merits of knowledge generalization, but also the ability of knowledge extension. In fault diagnosis of rotating machinery based on the acquired rules, a flexible matching model between new objects and rules is established, which makes the diagnostic conclusion be able in accordance with different requests.