基于粗糙集与神经网络集成的内燃机故障诊断

Fault Diagnosis Based on Integration of Rough Sets and Neural Networks

鉴于粗糙集理论对于决策系统的约简处理能力以及神经网络的自组织聚类和非线性映射功能,提出了应用SOM网络—粗糙集—BP网络集成进行故障诊断的方案:应用SOM网络离散化故障诊断数据中的连续属性值;基于粗糙集理论计算诊断决策系统的约简,根据实际需要确定最优决策系统;在最优决策系统的基础上设计BP网络进行故障诊断。柴油机的实际诊断结果验证了将神经网络与粗糙集理论相结合进行故障诊断的可行性。在数据充分的条件下,该方案可以推广应用于其它机械设备。

Considering the ability of rough sets theory to reduction of decision system and that of neural networks to clustering and nonlinear mapping,a new hybrid system of rough sets and neural networks for intelligent fault diagnosis was presented. Firstly, the continuous attributes in diagnostic decision system were discretized with selforganizing mapping neural network. Then, reducts were found based on rough sets theory, and the optimal diagnostic decision system was determined. Lastly, according to the optimal decision system, BP neural classifier was designed for fault diagnosis. The diagnosis of a diesel demonstrated that the solution can reduce the cost of diagnosis and increase the efficiency of diagnosis, while preserving the precision, and verified the feasibility of the engineering application. With enough sample data, the solution can be applied to other machinery.