基于粗糙集神经网络的车载设备故障诊断

Diagnosis on Failure of Onboard Equipment Based on Rough Set Neural Network

由于神经网络在车载设备故障诊断中存在网络结构复杂、训练时间长的问题,利用粗糙集理论处理不确定数据的优势,对BP神经网络进行优化。提出属性约简算法,去掉冗余信息,保留必要属性。通过对现场车载设备故障数据的实例分析表明,优化前后性能提升明显,且使用该方法能有效减少输入层神经元个数,提高车载设备故障诊断的效率和准确度。

Since the neural network has complicated structure and requires comparatively long time for diagnosis of onboard device failure, the advantage of rough set in processing uncertain data is leveraged to optimize the BP neural network. Attribute reduction algorithm is proposed to eliminate the redundant information and keep the necessary attributes. Analysis on actual data of onboard device failures has shown that the performance has been improved apparently before and after the optimization. Moreover, the method could well reduce the number of neurons at the input layer and improve the efficiency and accuracy of onboard device failure diagnosis.