Improved delignification kinetics models through AI-handling of data from large scale processes

Purpose and goal:

The purpose of the project is to develop improved experimental methods and kinetic models to study and design kraft pulping process by using large scale process data for AI-driven models. Specific aims are to: - Improve experimental design through AI-based identification of critical experimental relations

- Develop machine learning methods for: black-box end-to-end generation of comparable kinetic models based on experimental data & gray-box extended kinetic models combing existing domain knowledge with data-driven learning, aiming at a white-box models directly from data. Expected results and effects:

The project is expected to deliver following results:

- Identification of critical experimental conditions during kraft delignification through AI-treatment of large-scale process data

- AI methods for development of data-driven kinetic model and those combing existing knowledge with data-driven learning

Implementation of these results will directly contribute to improved process assessment, research approaches and new insights pertaining to process control & design, leading to increased resource efficiency, process flexibility and sustainability of the kraft pulping. Approach and implementation:

The project will be structured in two work packages (two post doc projects, 24 months each), addressing the problem from two different directions each: AI and delignification technology. The AI work package, WP1, (at Linköping University) will aim at developing a pure data driven model, while the delignification technology postdoc, WP2, (at Chalmers) will focus on gradually improving the existing analytical models with data-driven components.

These two efforts will towards the end of the project converge into joint efforts towards developing an improved white-box model.