Batch-to-batch optimization of cobalt oxalate synthesis process using modifier-adaptation strategy with latent variable model

Abstract

A batch-to-batch optimization strategy is proposed to cope with the presence of data-driven model uncertainty, and applied to a cobalt oxalate synthesis process. In order to overcome the difficulties in developing the first principle model of a fed-batch synthesis process, latent variable model incorporated with batch-wise unfolding approach is used to describe the causal relationship between the manipulated variables and the final product quality. Based on modifier-adaptation methodology, the idea of using final quality measurements in response to handling data-driven model uncertainty is formalized. And the batch-to-batch update criterion is presented to compensate for the mismatch of the necessary condition of optimality (NCO) between the plant and the data-driven model based optimization problem. The extensions to avoid using the latent variable model for extrapolation and to model the variability in initial conditions are also presented. Simulation studies demonstrate that the proposed strategy can result in better optimization performances compared to traditional model-adaptation strategy from batch to batch.