(576g) A Unified Approach for Identifying Dynamic Models Together with Terminal Quality Models for for Batch Processes

Batch processes are commonly found in many domains including chemical, mechanical, biochemical, agriculture and pharmaceutical industries due to the requirement of producing high-value products. Since the productions amount is small in batch process, it is important to maintain consistency in obtaining products with excellent quality and this can be only achieved by deploying a suitable control strategy. Production can be made profitable by using predictive control (MPC) to significantly reduce the waste costs and in order to develop such a strategy one must focus on the modelling aspect associated with the same. Typically in batch processes, there are dynamic variables which are measured at every time step and a set of quality variables associated with the final product, whose measurements are only available after the end of a batch. A good modeling strategy needs to incorporate both the dynamic and the quality information to fully utilize the available data. Various techniques have been developed over the years to address quality control in batch processes [1] and the most recent work [2] focuses on constructing a Linear Time Invariant State Space (LTI SS) model for capturing the dynamics of the process and a Partial Least Squares (PLS) regression model to establish the connection between the terminal dynamic "states" of a batch given by the LTI SS model with the quality measurements of that particular batch. This strategy has overcome the challenges posed using previous techniques like dealing with uneven batch lengths and missing data, and also has achieved success in achieving closed loop control in a batch Rotational molding process [3]. However, these models were identified separately, with LTI SS model being identified first using subspace identification, followed by the PLS quality model. This might result in a suboptimal set of models, since the identification procedure for the dynamic model doesn't take into account the quality measurements. To ensure that both the models are connected during the identification, this work introduces an optimization framework designed to jointly identify the dynamic and the quality models for batch processes. The performance of this integrated model will be demonstrated using real experimental data obtained from a Uni-axial Rotational Molding process, which serves as a representative example of a typical batch process, and also will be compared with the performance of the previous approach.

References:

[1] Flores-Cerrillo J, MacGregor J. Control of batch product quality by trajectory manipulation using latent variable models. J Process Control. 2004;14:539–553

[2] Corbett, Brandon, and Prashant Mhaskar. "Subspace identification for data‐driven modeling and quality control of batch processes." AIChE Journal 62.5 (2016): 1581-1601.

[3] Garg, Abhinav, et al. "Model predictive control of uni-axial rotational molding process." Computers & Chemical Engineering 121 (2019): 306-316.