For dynamic process modeling and monitoring purpose, it is desirable to extract both the auto-correlations and the cross-correlations in measurements. Besides, the proposed dynamic model is also expected to provide an accurate prediction, visualization and an explicit interpretation of the data structures. From this perspective, the dynamic latent variable model is more suitable compared to the traditional dynamic principal component analysis (DPCA). In this paper, a novel independent dynamic latent variable model is proposed to explicitly extract several independent dynamic latent variables with which to capture process dynamics in the measurements. The proposed model is derived in the probabilistic framework and the model parameters are estimated via the expectation-maximum algorithm. Finally, a case study is illustrated to evaluate the performance of the proposed method for dynamic modeling and process monitoring.
|Title of host publication||Proceedings of 2019 11th CAA Symposium on Fault Detection, Supervision, and Safety for Technical Processes, SAFEPROCESS 2019|
|Publisher||Institute of Electrical and Electronics Engineers|
|Number of pages||7|
|Publication status||Published - Jul 2019|
|Event||11th CAA Symposium on Fault Detection, Supervision, and Safety for Technical Processes (SAFEPROCESS 2019) - Xiamen, China|
Duration: 5 Jul 2019 → 7 Jul 2019
|Conference||11th CAA Symposium on Fault Detection, Supervision, and Safety for Technical Processes (SAFEPROCESS 2019)|
|Period||5/07/19 → 7/07/19|
Bibliographical noteThis work is supported by the National Natural Science Foundation of China (61603342), NSFC-Zhejiang Joint Fund for the Integration of Industrialization and Informatization (U1609214) and Zhejiang Provincial Natural Science Foundation of China (LY19F030003).
- Dynamic modelling
- Independent Dynamic Latent Variable
- Process Monitoring