旋转机械故障诊断增量式属性约减新方法

New Incremental Attribute Reduction Method of Rotating Machinery for Fault Diagnosis

粗糙集理论是旋转机械属性约减的常用工具,但无法有效用于样本逐渐增多的情况。针对增量式属性约减问题,提出了基于二进制分辨矩阵的属性约减算法。将新增对象分为四种类型,分别介绍了相应分辨矩阵的更新情况,从而得到相应的约减结果。该算法对于增量不需要重复计算整个分辨矩阵。通过和其他算法比较可知该算法可以快速得到约减结果,并具有较小的空间复杂度。最后,利用实际滚动轴承数据进行故障诊断实验,验证了所提算法的正确性。

Rough set theory is a useful tool for attribute reduction of fault diagnosis for rotating machinery, but cannot be efficiently used to sample increased areas. Aiming at the problem of incremental attribute reduction, the attribute reduction algorithm was put forward based on the binary resolution matrix. The new incremental objects for reduction were divided into four types and respectively introduced the updates of the discernibility matrix for different situations, to obtain the corresponding reduction results. In the proposed algorithm, the whole resolution matrix did not need to calculate repeatedly. Compared with other algorithms ,it can quickly get the reduction results, and has a smaller space complexity. Finally, the fault diagnosis experiment was made by using the actual data of rolling bearing, which verified the correctness of the mentioned method.