基于堆栈自动编码器的泵站机组故障分析

将堆栈自动编码器(Stack Auto-encoders)应用到泵站机组的故障分析中,构建了基于堆栈自动编码器的故障分析模型。构建的模型主要由输入层、3个中间隐层和输出层构成,以实现对泵站机组的监测数据和特征进行提取和降维处理。模型网络采用了非监督逐层贪婪方法训练,然后使用反向传播算法对网络参数予以优化,在此基础上,利用softmax分类器进行分类。实验结果表明,运用所构建的模型对机组故障以及不同工况的平均分类准确率可以达到79.88%。该成果可以为泵站机组故障分析提供一定的参考依据。

Failure analysis for pumping stations units based on Stack Auto-encoders

Stack Auto-encoders is applied to the failure diagnosis of pumping stations units, based on which an diagnosis model consisting of 5 layers, including a input layer, 3 middle layers and an output layer is constructed to realize feature extraction and dimension reduction of pumping unit data. The Greedy Layer -Wise Unsupervised Learning Algorithm is used to train each layer, back propagation algorithm is applied to optimize the parameters of the network, and softmax classifier is used to the data classification. The results show that the average classification accuracy of pump unit failure is 79.88% on different working conditions. This study can provide references for failure analysis of pumping stations.