.png){kind=icon}
.png){kind=icon}
Return
A hybrid causal-inference and neuro-fuzzy framework for advanced process monitoring
.png){kind=icon}
PRE
AI
H
J
F
DOI：10.1016/j.jprocont.2026.103724.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The emergence of Industry 5.0, automation, and advanced sensor networks has made modern industrial processes increasingly intricate and dynamic. Traditional monitoring methods often fall short in addressing the demands of real-time anomaly detection, key variable identification, and root cause analysis in chemical and industrial processes. To overcome these challenges, this work introduces a hybrid causal-inference and neuro-fuzzy framework that integrates dynamic inner global–local preserving projection (DiGLPP), an adaptive neuro-fuzzy inference system (ANFIS), and a causal lineage graph (CLG). The framework is designed to detect faults, highlight influential variables, and trace the propagation paths of causal faults across process units. Its performance is benchmarked against established models, such as Wavelet-Principal Component Analysis (Wavelet-PCA), Dynamic PCA (DPCA), and Dual-Attention Long Short-Term Memory Autoencoder (DALSTM-AE), with the Tennessee Eastman Process (TEP) serving as the primary benchmark. Results demonstrate that the proposed methodology not only improves the detection of challenging fault scenarios in the TEP but also enables robust anomaly detection, accurate fault classification and isolation, identification of root-cause channels, and diagnosis of critical process variables. These findings highlight its potential to meet the practical monitoring needs of complex real-world industrial systems.
Keywords:
Industry 5.0
fault detection
causal inference
neuro-fuzzy system
process monitoring
Journal
IF:
3.9
Papers:
3.4K
Citations:
7.3K
Organization
Cited Papers