Data-Driven Modeling of Sequential Batch-Continuous Process

Data-Driven Modeling of Sequential Batch-Continuous Process

Jungup Park, Michael Baldea, Thomas F. Edgar

Department of Chemical Engineering

The University of Texas at Austin, 1 University Station C0400, Austin, TX 78712

Email: mbaldea@che.utexas.edu

Complex chemical processes can involve both batch and continuous stages. In this case, raw materials and other ingredients are initially processed batch-wise, prior to being fed to a processing line that operates continuously. The operating conditions and operating performance of both the batch and the continuous stages have an impact on the final product quality.

Such batch-to-continuous processes pose specific analysis and control challenges. The batch side of the process operation is carried out periodically at specified time intervals. After each operating instance, the batch product is fed to the continuous production flux. Empirical evidence suggests that this mode of operation leads to a deterioration of the causal relation between the properties of the batch product and the quality of the product of the continuous process. This is further complicated by the time delay that is inherently introduced by the continuous stage of the process between the completion of the batch stage and any quality measurements obtained from the final product.

In this contribution, we focus on using data-driven modeling tools to define a framework for establishing causality between batch properties and continuous process. Subsequently, we will focus on the batch monitoring, correlating batch to continuous data, and modeling the end quality of the product using different set of training data.