An Industrial Fault Diagnosis Method Based on Graph Attention Network DOI
Yan Hou, Jinggao Sun, Xing Liu

и другие.

Industrial & Engineering Chemistry Research, Год журнала: 2024, Номер 63(44), С. 19051 - 19062

Опубликована: Окт. 22, 2024

In the field of industrial production, precise and timely implementation fault diagnosis methods is crucial for improving product quality, enhancing operational safety, reducing downtime, minimizing losses. Recent studies have shown that most CNN-based models are more suitable handling Euclidean data such as images or videos but not dealing with non-Euclidean sensor data. practical scenarios, chemical process imbalanced patterns may lead data-driven to assign different attentions patterns. The SMOTE algorithm commonly used generate new data, it often tends overfit when there very few nearest neighbor samples. To address these issues, we designed an efficient model named KRGAT. fully utilize spatial structural information on employed graph attention networks (GATs), which well-suited Additionally, introduced top-k loss method select hard samples, thereby increasing weight Furthermore, improved DropMessage enhance model's accuracy robustness. Experimental results demonstrate our outperforms baseline under both balanced conditions.

Язык: Английский

Bidimensionally partitioned online sequential broad learning system for large-scale data stream modeling DOI Creative Commons
Wei Guo,

Jianjiang Yu,

Caigen Zhou

и другие.

Scientific Reports, Год журнала: 2024, Номер 14(1)

Опубликована: Дек. 30, 2024

Incremental broad learning system (IBLS) is an effective and efficient incremental method based on paradigm. Owing to its streamlined network architecture flexible dynamic update scheme, IBLS can achieve rapid reconstruction the basis of previous model without entire retraining from scratch, which enables it adept at handling streaming data. However, two prominent deficiencies still persist in constrain further promotion large-scale data stream scenarios. Firstly, needs retain all historical perform associated calculations process, causes computational overhead storage burden increase over time as such puts efficacy algorithm risk for massive or unlimited streams. Additionally, due random generation rule hidden nodes, generally necessitates a large size guarantee approximation accuracy, resulting high-dimensional matrix calculation poses greater challenge updating efficiency model. To address these issues, we propose novel bidimensionally partitioned online sequential (BPOSBLS) this paper. The core idea BPOSBLS partition feature aspects instance dimension dimension, consequently decompose least squares problem into multiple smaller ones, then be solved individually. By doing so, scale complexity original high-order are substantially diminished, thus significantly improving usability complex tasks. Meanwhile, ingenious recursive computation called devised solve BPOSBLS. This exclusively utilizes current samples iterative updating, while disregarding samples, thereby rendering lightweight with consistently low costs requirements. Theoretical analyses simulation experiments demonstrate effectiveness superiority proposed algorithm.

Язык: Английский

Процитировано

1

System identification of a nonlinear continuously stirred tank reactor using fractional neural network DOI Creative Commons
Meshach Kumar, Utkal Mehta, Giansalvo Cirrincione

и другие.

South African Journal of Chemical Engineering, Год журнала: 2024, Номер 50, С. 299 - 310

Опубликована: Сен. 19, 2024

Язык: Английский

Процитировано

0

Adaptive soft sensor using stacking approximate kernel based BLS for batch processes DOI Creative Commons
Jinlong Zhao, Mingyi Yang, Zhigang Xu

и другие.

Scientific Reports, Год журнала: 2024, Номер 14(1)

Опубликована: Июнь 4, 2024

Abstract To deal with the highly nonlinear and time-varying characteristics of Batch Process, a model named adaptive stacking approximate kernel based broad learning system is proposed in this paper. This innovatively introduces (AKBLS) algorithm Adaptive Stacking framework, giving it strong fitting ability, excellent generalization ability. The Broad Learning System (BLS) known for its shorter training time effective processing, but uncertainty brought by double random mapping results poor resistance to noisy data unpredictable impact on performance. address issue, paper proposes an AKBLS that reduces uncertainty, eliminates redundant features, improves prediction accuracy projecting feature nodes into space. It also significantly computation matrix searching kernels enhance ability industrial online applications. Extensive comparative experiments various public datasets different sizes validate this. framework utilizes ensemble method, which integrates predictions from multiple models using meta-learner improve generalization. Additionally, employing moving window method—where fixed-length slides through database over time—the gains allowing better respond gradual changes Process. Experiments substantial dataset penicillin simulations demonstrate predictive compared other common algorithms.

Язык: Английский

Процитировано

0

Cloud Regularized Stochastic Configuration Network for Few-shot Fault Diagnosis of Process Industries DOI
Xiaoqi Wang, Jiang Liu, Lanhao Wang

и другие.

2022 34th Chinese Control and Decision Conference (CCDC), Год журнала: 2024, Номер unknown, С. 70 - 75

Опубликована: Май 25, 2024

Язык: Английский

Процитировано

0

An Industrial Fault Diagnosis Method Based on Graph Attention Network DOI
Yan Hou, Jinggao Sun, Xing Liu

и другие.

Industrial & Engineering Chemistry Research, Год журнала: 2024, Номер 63(44), С. 19051 - 19062

Опубликована: Окт. 22, 2024

In the field of industrial production, precise and timely implementation fault diagnosis methods is crucial for improving product quality, enhancing operational safety, reducing downtime, minimizing losses. Recent studies have shown that most CNN-based models are more suitable handling Euclidean data such as images or videos but not dealing with non-Euclidean sensor data. practical scenarios, chemical process imbalanced patterns may lead data-driven to assign different attentions patterns. The SMOTE algorithm commonly used generate new data, it often tends overfit when there very few nearest neighbor samples. To address these issues, we designed an efficient model named KRGAT. fully utilize spatial structural information on employed graph attention networks (GATs), which well-suited Additionally, introduced top-k loss method select hard samples, thereby increasing weight Furthermore, improved DropMessage enhance model's accuracy robustness. Experimental results demonstrate our outperforms baseline under both balanced conditions.

Язык: Английский

Процитировано

0