粗糙集-神经网络故障诊断方法研究

Research on Rough Set-Neural Network Fault Diagnosis Method

从人工神经网络故障诊断模型的特点出发 ,利用粗糙集理论解决该模型应用中的主要问题 ,包括进行训练样本质量研究 ,定义相关概念 ,给出故障特征提取算法等 ,提出了粗糙集 神经网络智能混合系统模型 ,分析了该模型的实现步骤 ,结合具体实例验证了上述理论的正确性·利用SAS软件进行了数值仿真·结果表明 ,提出的理论较好地解决了神经网络结构、训练样本的大小、样本质量等对人工神经网络的精度及泛化能力有直接影响的问题 ,减少了训练所需的计算量和时间 ,提高了模型的正确率·

The rough set theory was used to study training sample quality,define relative concept and establish fault feature extraction algorithm for Artificial Neural Network fault diagnosis model. An intelligent rough set neural network hybrid system model was brought forward. The realization steps of the model were analyzed. The validity of these methods was tested by practical examples. Simulation was done using SAS software. The method can solve ANN architecture, sample size,and sample quality,decrease the computation time,and increase the diagnosis correctness.