基于粗糙神经网络的传感器网络故障诊断

摘要

An intelligent fault diagnosis method combining the rough set and the artifieial neural network ( RS-ANNE) is presented to solve the difficulties in wireless sensor network faidt diagnosis caused by energy limitation and substantive superfluous information. It first utilises the attribute reduction of rough set theory to extract the minimum fault diagnosis feature set which contributes the most to fault diagnosis, and then determines the preliminary topological structure of neural network according to minimum fault diagnosis feature, followed by setting up the mapping relation between the fault features and the faults themselves, in the end the final diagnosis results are derived through subnet voting. Experimental results show that the diagnosis accuracy of RS-ANNE method is 95. 67% , it has 22. 98% less computation load and 13.88% higher diagnosis accuracy than those of the ANNE' s.%为了克服大量信息冗余和能量有限给无线传感器网络故障诊断带来的困难,提出一种将粗糙集与神经网络集成相结合的智能故障诊断方法(RS-ANNE).该方法首先利用粗糙集理论的属性约简技术,提取诊断故障贡献最大的最小故障诊断特征集合,然后根据最小故障诊断特征确定神经网络的初始拓扑结构,建立故障特征与故障之间的映射关系,最后通过子网表决得到最终诊断结果.实验结果表明,RS-ANNE诊断方法诊断正确率为95.67％,与ANNE方法相比计算量减小22.98％,诊断正确率提高13.88％.