(364p) Machine Learning Aided Tools for Separation Processes

Throughout my research journey, I have been focusing on the development and application of computational Process Systems Engineering (PSE) and machine learning methods to various separation systems in the process industry.

Research Interests:

The focus of my PhD is on the development of computational tools to aid in process design, optimisation and control of dynamic systems. More specifically, my work is dedicated to creating a comprehensive methodology for developing dynamic machine learning-based models for continuous separation processes. The goal is to facilitate their application in online optimisation and control, thereby improving efficiency and adaptability in industrial settings. The methodology allows for the development of dynamic data-driven and hybrid models of complex separation systems, by combining physicochemical knowledge of the processes with data training. The main advantage of this approach is in the direct prediction of the systems’ dynamic profile based on the predefined operating conditions of the process. As such, model complexity is greatly reduced, and the developed models can be used for further model-based applications, such as online optimisation and control.

The methodology has been successfully implemented on a multi-column chromatographic separation process, used for the purification of monoclonal antibodies. Both hybrid and data-driven models of the system have been developed and validated against an experimentally validated high-fidelity model of the process. They have also been successfully implemented in a process optimisation framework, showcasing the models’ suitability for online monitoring of the process. To further the applicability of this methodology, current research is focused on two primary areas. The first area involves investigating the performance of the hybrid model across different products, through transfer learning. The second area pertains to the development of a control scheme using the data-driven models of the process. This control scheme leverages reinforcement learning (RL) to continuously optimise decision-making in real-time. By integrating RL, the system can dynamically adapt to changes in the environment, improve operational efficiency, and maintain optimal performance under varying conditions.

Based on the knowledge and skills that I have acquired through my research journey, I am looking for an industry position where I can apply my technical skills and collaborate with a team to solve real-world problems. I am eager to learn and contribute to an organisation that values innovation and offers opportunities for professional growth.