基于S-RNPAE算法的间歇过程早期故障监测

扫码查看

原文链接

NETL
NSTL
万方数据

中文摘要：针对具有多变量、非线性和高维度特点的间歇过程数据使得早期故障信号易被噪声干扰且故障幅值低导致故障监测效果不佳的问题，提出一种基于堆叠鲁棒邻域保持自编码(stack-robust neighborhood preserving autoencoder，S-RNPAE)的间歇过程早期故障监测方法。首先，通过L2。1范数重新设计自编码器的目标函数，以提高模型对噪声和离群点的鲁棒性;其次，利用邻域保持嵌入来正则化鲁棒自编码器的方式构建鲁棒邻域保持自编码(robust neighborhood preserving autoencoder，RNPAE)模块，解决自编码器作为一种全局模型而忽略包含早期故障特征的局部近邻信息的提取问题;然后，将多个RNPAE模块堆叠构造S-RNPAE网络，从而获取深层全局-局部特征，保证对早期微小故障信息提取更充分，并建立检测统计量实现过程检测;最后，利用一种适用于非线性过程的贡献图方法完成故障诊断，其诊断结果更准确。通过Swiss Roll数据集和青霉素发酵过程的实验表明，所提方法的特征提取能力更强，对间歇过程的早期故障更敏感，具有更好的早期故障监测效果。

外文标题：Incipient fault monitoring of batch process based on S-RNPAE algorithm

外文摘要：For batch process data with multivariate,nonlinear,and high dimensional characteristics,the early fault signals are easy to be disturbed by noise and the fault amplitude is low,which lead to poor fault monitoring performance,we propose an early fault detection method for batch processes based on stack-robust neighbourhood preserving autoencoder(S-RNPAE)in this paper.Firstly,the objective function of the autoencoder is redesigned by the L2,1 norm to improve the model robustness to noise and outliers.Secondly,the robust neighbourhood preserving autoencoder(RNPAE)module is constructed by using neighbourhood preserving embedding to regularize the robust autoencoder,which solves the problem of the autoencoder as a global model that ignores the local neighbour information containing early fault features.Then,we construct the S-RNPAE network by stacking multiple RNPAE modules,which can obtain deep global-local features to ensure the extraction of early minor fault information more fully,and we establish detection statistics to realize process monitoring.Finally,the nonlinear contribution diagram method is used to complete the fault diagnosis,which is more accurate for nonlinear variables.The examples of the Swiss Roll dataset and the simulation process of penicillin fermentation show that the proposed method has stronger feature extraction ability,which is more sensitive to the early fault of batch process and has a better early fault monitoring performance.

外文关键词：

batch processincipient faultprocess monitoringautoencoderneighborhood preserving embedding

作者：

刘凯、赵小强、牟淼、张妍

展开 >

作者单位：

兰州理工大学电气工程与信息工程学院,兰州 730050

甘肃省工业过程先进控制重点实验室,兰州 730050

兰州理工大学国家级电气与控制工程实验教学中心,兰州 730050

关键词：

间歇过程早期故障过程监测自编码器邻域保持嵌入

基金：

国家重点研发计划国家自然科学基金国家自然科学基金甘肃省科技计划甘肃省科技计划甘肃省工业过程先进控制重点实验室开放基金甘肃省教育厅产业支撑项目

项目编号：

2020YFB1713600622630216216302321JR7RA20621YF5GA0722022KX072021CYZC-02

出版年：

2024

DOI：

10.13195/j.kzyjc.2022.1401

控制与决策

东北大学

控制与决策

CSTPCD北大核心

影响因子：1.227

ISSN：1001-0920

年,卷(期)：2024.39(5)

参考文献量32