Semi-supervised fault classification method based on weighted feature alignment auto-encoder

Abstract

The invention discloses a semi-supervised fault classification method based on a weighted feature alignment auto-encoder, and the method comprises the steps: carrying out the reconstruction pre-training of a stacked auto-encoder through employing labeled data, and estimating the probability density distribution of a reconstruction error; then, according to the probability density function of the training data reconstruction error, calculating the weight of the unlabeled sample; and furthermore, constructing a semi-supervised classification model based on a weighted feature alignment auto-encoder by using a labeled sample set, an unlabeled sample set and corresponding weights. The weighted feature alignment auto-encoder classification model designs a cross entropy training loss function based on a weighted Sinkhorn distance, and the function enables the model to use label data and label-free data at the same time in a fine tuning stage, so that not only can deep mining of data information be realized, but also the generalization ability of the network model can be improved. Meanwhile, due to the introduction of a weighting strategy, the robustness of the model is remarkably improved.

Description

technical field [0001] The invention belongs to the field of industrial process control, in particular to a semi-supervised fault classification method based on weighted feature alignment autoencoders. Background technique [0002] Modern industrial processes are developing toward large scale and complexity. How to ensure the safety of the production process is one of the key issues that need to be solved in the field of industrial process control. Fault diagnosis is a key technology to ensure the safe operation of industrial processes, and it is of great significance to improve product quality and production efficiency. Fault classification is a part of fault diagnosis. By learning from historical fault information, automatic identification and judgment of fault types can be realized, thereby helping production personnel to quickly locate and repair faults and avoid further losses caused by faults. With the continuous development and progress of modern measurement methods...

AI Technical Summary

Problems solved by technology

However, this assumption has its limitations. The data collected by industrial processes often contain a lot of noise and abnormal points, and the drift of working conditions may occur. Labeled data are often manually screened and marked by experts in the process field. , while the unlabeled samples have not been screened, therefore, it is very likely that there will be abnormal data in the unlabeled data that are different from the distribution of the labeled data

When the distribution of unlabeled data is not consistent with that of labeled data, the performance of semi-supervised algorithms will drop, even lower than that of supervised algorithms that only use labeled data for training.

Images