Stage-based Multiple PCA Modeling and On-line Monitoring Strategy for Batch Processes

The MPCA model takes the entire batch data as a single object, and it is difficult to reveal the changes of process correlation from stage to stage. Considering that the multiplicity of the operation stage is an inherent nature of many batch processes, a stage-based model was developed by the clustering algorithm based on the process correlation characteristic. However, misclassification may occur at the beginning and end of each stage, because the k-means clustering algorithm is a hard-partition method in dealing with patterns between two neighboring clusters. To resolve the above matters flexibly, multiple PCA modeling based on the soft partition algorithm was introduced. It reduces the false alarm and missing alarm for batch process in on-line monitoring due to batch variation. The application to three-tank plant experiment demonstrates the effectiveness of the method