【摘要】 针对新型固体燃料电池(SOFC)系统在多工况下快速识别故障的需要,研究基于深度自动编码器(AutoEncoder)的故障特征提取方法,通过处理多种工况下的工业数据,提取出有效的特征。并使用基于Softmax多层神经网络对特征数据进行模式分类,达到多故障识别的目的。通过实践证明,基于深度神经网络和多层分类网络的故障识别系统能够有效地识别多工况下的不同故障。更多还原

【Abstract】 Considering the demand for rapid fault identification of new solid oxide fuel cell( SOFC) system under multiple operating conditions,the Auto Encoder-based fault feature extraction method was studied and it can process industrial data under a variety of operating conditions to get effective characteristics. And based on Softmax multi-layer neural network,the feature data can be classified into patterns to achieve multi-fault identification. The practice proves that this fault identification system based on deep neural network and multi-layer classification network can effectively identify different faults under multiple operating conditions.更多还原