The mechanism of classification of the RNNs was revealed in this study. The benchmark of the Tennessee Eastman process was used to demonstrate the performance of the RNN-based fault diagnosis model. This study investigated the fault diagnosis performance of the entire and early detections for complicated chemical processes using the time-series recurrent neural networks (RNNs). The investigation included various layers and neuron nodes in RNNs using lean and rich training datasets and compared these RNNs with the artificial neural networks (ANNs). The results showed that the RNNs had better classification accuracies than the ANNs regardless of using lean or rich training datasets. The general classification mechanism was a priori classification that used normal operating data as the center so that it was incapable of separating the fault types having similar features of the normal operating data. RNNs drove the normal operating data out of the center and created the even spatial distribution of fault types, leading that RNNs were effective in classifying the fault types with subtle features when there is sufficient data.

这项研究揭示了RNNs的分类机制。田纳西州伊士曼过程的基准用来证明基于RNN的故障诊断模型的性能。这项研究使用时间序列递归神经网络（RNN）研究了复杂化学过程的整体和早期检测的故障诊断性能。该研究使用精益和丰富的训练数据集对RNN中的各个层和神经元节点进行了比较，并将这些RNN与人工神经网络（ANN）进行了比较。结果表明，无论使用精简或丰富训练数据集，RNN都比ANN具有更好的分类准确性。通用分类机制是一种以正常运行数据为中心的先验分类，因此它无法分离出具有正常运行数据特征的故障类型。RNN将正常的运行数据移出中心，并创建了故障类型的均匀空间分布，这导致当有足够的数据时，RNN可以有效地对具有细微特征的故障类型进行分类。