基于粗糙神经网络的故障诊断技术研究

Research of Fault Diagnosis Based on Rough Neural Network

针对故障诊断中设备监控数据越来越多的特点,提出用于故障诊断的粗糙神经网络模型。此模型的创新点是基于SOFM网络和差别矩阵的离散化算法,此算法不但指导属性划分类数,而且保证了得到最优属性约简,同时,充分利用了粗糙集和神经网络的故障诊断能力来保证诊断结果的准确性和彻底性。实践证明:此模型在工程上有着很好的适用性和可信性,能够为解决现代工业工程中的故障诊断提供有效的参考。

Because of more datas produced from equipment condition monitoring, a rough neural network model used in fault diagnosis is povided. The innovation of this model is the discrete algorithm based on SOFM network and discernibility matrix, this algorithm doesn＇t only guide dividing the number of attributes＇ category, but also guarantee getting optimal attribute reduction, meanwhile, the diagnosis ability of routh set and neural network can ensure the diagnosis accuracy and completeness. Practice indicates this model has good engineering applicability and creditability and can provide effective reference for fault diagnosis.