Phase Identification and Online Monitoring for the Uneven Batch Processes

• Published in: IEEE access Vol. 7; pp. 81351 - 81363
• Main Authors: Guo, Runxia, Jin, Yancheng
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 2019; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Aircraft; Algorithms; Batch processes; Batch processing; Batch production systems; Correlation; Covariance matrices; Covariance matrix; Data models; Fault diagnosis; Feature extraction; fine modeling; Monitoring; moving window-based multiway information increment matrix (MWMIIM); online monitoring; Parameter identification; Partitioning algorithms; Partitions; Phase identification; Steering; uneven batch processes
• ISSN: 2169-3536; 2169-3536
• DOI: 10.1109/ACCESS.2019.2919167

In practice, the batch processes are usually uneven and show significantly different variables' characteristics in different sub-phases. Therefore, it is necessary to divide each batch into several sub-phases separately. In this paper, a moving window-based multiway information increment matrix (MWMIIM) algorithm for the uneven batch processes is proposed for single-batch phase identification and online monitoring by combining the moving window technique with an information increment matrix (IIM) algorithm. Similar to the IIM algorithm, the MWMIIM algorithm captures variables' correlation changes by calculating the increment matrix between adjacent covariance matrices, which does not need to extract feature from data or carry out complicated matrix decomposition, thus improving the computational efficiency. Besides, the influence of several vital parameters on the phase identification performance is discussed in detail. After phase identification, the partition points of each sub-phase need not to be strictly aligned. Furthermore, a batch process is divided into three types of regions, namely common region, transition region, and end region. Next, fine modeling and online monitoring strategies are adopted in different regions separately. The comparative experiment is conducted by the window-based stepwise sequential phase partition method for nonlinear uneven batch processes (WNSSPP-U). A practical application on batch processes, namely aircraft steering gear system fault diagnosis experiment, is given to confirm the feasibility and effectiveness of the proposed method.