Data Synchronization and Recursive Optimization of Uneven-length Batch Processes

We propose a data synchronization approach based on the characteristic trajectory of multistage fusion to solve the problem of uneven-length data in batch process optimization, and then we optimize the operation trajectory recursively based on the index increment and the loading cosine similarity. In this method, we first use principal component analysis (PCA) to project the variation features of the data along the time dimension into a scoring space and obtain the characteristic trajectory variation based on the intrinsic time. Considering the multistage characteristics of batch processes, we apply the K-means algorithm to perform phase segmentation of the data, and synchronize and fuse the data based on the feature differences at each stage to realize batch data synchronization of the multistage fusion. Then, using the synchronized data, we optimize the operation trajectory recursively based on the index increment and the loading cosine similarity. We apply the proposed method to the batch crystallization process of a chemical product and found the simulation results to confirm its effectiveness and advantages.