1
On-line detection of homogeneous operation ranges by dynamic principal component analysis based time-series segmentation
László Dobos, János Abonyi*
Corresponding author contact information
Department of Process Engineering, Faculty of Engineering, University of Pannonia, 8200 Veszprem, Hungary
abonyij@fmt.uni-pannon.hu
Abstract
Development of chemical process technologies shall be based on the analysis of process data.
In the field of process monitoring the recursive Principal Component Analysis (PCA) is widely applied to detect any misbehavior of the technology. The investigation of transient states needs dynamic PCA to describe the dynamic behavior more accurately. By combining and integrating the recursive and dynamic PCA into time series segmentation techniques, efficient multivariate segmentation methods were resulted to detect homogenous operation ranges based on process data. The similarity of time-series segments is evaluated based on the Krzanowski-similarity factor, which compares the hyperplanes determined by the PCA models. With the help of developed time series segmentation framework, separation of operation regimes becomes possible for supporting process monitoring and control. The performance of the proposed methodology is presented throughout a linear process and the commonly applied Tennessee Eastman process.
Highlights
► To improve the operational and control efficiency of a technology the detection of any faults and disturbances is necessary. ► The detection of disturbances needs an advanced, multivariate process monitoring and time-series segmentation algorithm. ► Novel Dynamic Principal Component Analysis based time-series segmentation framework is developed as process monitoring system. ► The developed framework is suitable to detect the disturbances that change the correlation structure of input-output data. ► The proposed methodology can be applied even for streaming and for historical process data.
Keywords
Time-series segmentation; Dynamic principlal component analysis; Variable forgetting factor;
Process monitoring; Data mining; Tennessee Eastman process
