State Estimation and Control of mAb Production Processes

• Author / Creator

  Obiri, Sandra Ama

• Monoclonal antibodies (mAbs) have become indispensable assets in modern medicine, necessitating advancements in large-scale production to meet the growing market demand. This thesis focuses on two critical aspects of mAb production: state estimation in the upstream process and the control and optimization of the downstream process.

  Accurate state estimation is vital for optimizing the mAb production process and reducing costs. Therefore, this work presents guidelines for sensor selection to enhance state estimation accuracy, and illustrates an effective variable selection technique for simultaneous state and parameter estimation in the upstream process. Subsequently, a Moving Horizon Estimation (MHE) framework is developed and applied to three case studies to demonstrate the efficiency of estimating some parameters in addition to the states, with the Root Mean Squared Error (RMSE) serving as the evaluation criterion.

  The switching of the downstream capture columns are pivotal for ensuring the continuity of the integrated continuous mAb production process. However, due to the discrete nature of the switching operation, advanced process control algorithms such as economic model predictive control (EMPC) are computationally difficult to implement.
  To address this issue, computationally-efficient approaches are explored to improve EMPC implementation. The first approach uses a sigmoid function to relax the discrete decision variables into continuous ones, which makes the optimization problem easier to solve. The second approach involves training a ReLU neural network to replace the original nonlinear model, leading to the conversion of the integer nonlinear program (INLP) into an integer linear program (ILP). This modification facilitates a quicker solution to the optimization problem.
  Additionally, we explore a reinforcement learning (RL) method, which seeks to identify the most effective policy for addressing the optimization problem.
  Comparatively, all three techniques are evaluated against the conventional switching approach, which relies on a fixed switching product breakthrough rule. The integration of improved state estimation in the upstream process and optimized control strategies in the downstream process presents a comprehensive framework for enhancing mAb production efficiency.

• Subjects / Keywords

  • monoclonal antibody production
  • economic model predictive control
  • rectified linear unit network
  • sigmoid function
  • reinforcement learning

• Graduation date

  Fall 2023

• Type of Item

  Thesis

• Degree

  Master of Science

• DOI

  https://doi.org/10.7939/r3-hdnz-a744

• License

  This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.