中文摘要：

针对滚动轴承,实现了一种粗糙集理论和神经网络技术相结合的新型的故障诊断虚拟系统.该系统利用粗糙集对知识的约简能力,对采集的故障征兆数据进行预处理,即采用竞争学习神经网络把连续属性离散化,将结果导入Rosetta软件中逐步分析处理,得到最小条件属性集,在此基础上构建BP神经网络进行故障识别,将网络输出送回LabView进行显示.实例分析表明,该系统可以提高滚动轴承故障诊断的收敛速度,在期望误差相同的情况下,网络训练时间减小了176步.

英文摘要：

Aiming at rolling bearings,the implementation procedure of a new style fault diagnostic system is presented in this paper.The combination of rough sets and BP neural network are adopted in the design of the diagnostic system.Utilizing the knowledge reduction ability of rough sets theory,the diagnostic system preprocesses the collected fault symptom data at first,i.e.the discretization of continuous attributes by using competition learning neural networks.The intermediate output is introduced to software of ＂Rosetta＂ to be analyzed step by step until the smallest condition attributed sets are obtained.Based on the smallest condition attributed sets,the BP networks are built,which are used to recognize the faults of rolling bearings and then transfer the fault states back to LabView for displaying.The example analysis indicated that the system can enhance fault diagnosis convergence speed and the network training time reduces 176 steps at the same expected error.