期刊
JOURNAL OF THE TAIWAN INSTITUTE OF CHEMICAL ENGINEERS
卷 132, 期 -, 页码 -出版社
ELSEVIER
DOI: 10.1016/j.jtice.2021.104200
关键词
Convolutional autoencoder; Deep learning; Fault-relevant feature; Fault diagnosis
资金
- National Natural Science Foun-dation of China [61773106, 61703086]
- Fundamental Research Funds for the Central Universities [N2104009]
A new feature learning method, SCAE, addresses the issue of CAE's inability to ensure that extracted features are related to fault types, by pretraining the network to learn internal spatial information and fault information. The fault-relevant features obtained by SCAE can clearly distinguish between different fault types, providing more appropriate predefined parameters for fine-tuning to enhance classification performance.
Background: Convolutional autoencoder (CAE) is an unsupervised feature learning method and shows excellent performance in multivariate fault diagnosis. However, CAE cannot guarantee that the extracted feature is always related to the fault type due to its unsupervised self-reconstruction in the pretraining phase.Methods: To solve this problem, a new feature learning method, supervised convolutional autoencoder (SCAE) is proposed to pretrain the network and learn representative feature containing internal spatial information and fault information. In the SCAE, process sample and corresponding label are reconstructed by multilayer encoding-decoding the raw sample. Meanwhile, to prevent label information overfitting the network, a minimum difference transformation function is introduced into the loss function.Findings: The obtained fault-relevant features can be obviously distinguished between different fault types. The trained pretraining network provides more appropriate predefined parameters for fine-tuning to improve the classification performance. The effectiveness of the proposed method is evaluated by the continuous stirred tank reactor (CSTR) process and the Tennessee Eastman (TE) process.(c) 2021 Taiwan Institute of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据