Abstract
Statistical process monitoring techniques, such as principal component analysis (PCA), have been widely used for monitoring complex industrial processes and diagnosing process and sensor faults. While the statistical methods are easy to apply in practice, they have difficulty handling dynamic processes and often lead to false alarms during dynamic transients. In this project, a subspace model based approach is developed to perform dynamic process monitoring and fault diagnosis. We make use of the subspace identification method (SIM) based on PCA (SIMPCA). Fault detection is accomplished using multivariate statistics on the parity space and its complemental space. Fault diagnosis is carried out with reconstruction-based contributions. The proposed method is compared against PCA using data from the Tennessee-Eastman Process (TEP). © 2012 IFAC.
Original language | English |
---|---|
Pages (from-to) | 684-689 |
Number of pages | 6 |
Journal | IFAC Proceedings Volumes |
Volume | 45 |
Issue number | 20 |
DOIs | |
Publication status | Published - Aug 2012 |
Externally published | Yes |
Event | 8th IFAC Symposium on Fault Detection, Supervision and Safety of Technical Processes, SAFEPROCESS 2012 - Mexico City, Mexico Duration: 29 Aug 2012 → 31 Aug 2012 |
Bibliographical note
ISBN: 9783902823090Keywords
- Fault diagnosis
- Parity space
- Process monitoring
- Reconstruction-based contributions
- Subspace identification
- Tennessee-Eastman Process