摘要
研究了一种基于粗糙集和禁忌神经网络(Tabu Search-Artificial neural network,TS-ANN)的故障诊断方法,解决了柴油机由于激振源多而导致的故障诊断困难的问题;首先通过SOM网络实现对初始决策表的属性值离散化,使用基于属性重要度的属性简约算法实现对决策表的属性简约,从而降低输入数据维数,然后通过禁忌算法实现对神经网络的隐层神经元个数以及权、阀值进行优化,将优化后的参数代入BP神经网络后进行训练以进一步调整,最后将训练好的神经网络用于实现故障诊断推理;仿真实验证明文中的方法能精确地实现故障诊断,且与其他方法相比,诊断精度分别提高了28.34%、13.45%和9.67%。
A fault diagnosis method based on rough set and TS ANN was researched to resolve the difficulty of diesel engine diagnosing due to too much excitation sources. Firstly, the SOM network was used to realize the discretization for attribute value, using the attribute re duction method based on the significance of attribute to simplify the attribute, then the tabu search algorism was given to optimize the neu rons of the hidden layer, the weight and the threshold and get the initial network, finally the initialed network is trained further to get the fi nal diagnosis model. The simulation result indicates the method in our paper can realize the fault diagnosis, and compared with the other method, Diagnostic accuracy were increased by 28.34%, 13.45 % and 9.67%.
出处
《计算机测量与控制》
北大核心
2013年第1期54-56,共3页
Computer Measurement &Control
