Physics-informed AI for identification and control of process-product relations in papermaking

Purpose and goal:

The project aims to evaluate AI techniques, such as physics-informed neural networks, which can integrate principles from well-known physical processes governing sheet formation, consolidation, and drying. By dynamically responding to paper manufacturing changes, AI-based tools can enhance process control. The project also aims to advance the knowledge of the paper and pulp industry on how these techniques can contribute to the industry´s role in shaping a future bio-based economy.

Expected results and effects:

The project aims to deliver a Proof-of-Concept for an analysis tool based on a self-learning model capable of predicting at least one observed variation in process and/or product. All results generated will be disseminated within the research school Resource-Efficient Processes and published if possible. The project´s impact should contribute to new opportunities for identifying and reducing resource efficiency gaps related to process variations in paper production.

Approach and implementation:

The project will be conducted over two years in close collaboration with two projects within the corporate research school Resource-Efficient Processes (RSP). As an initial step, data will be identified and collected from the collaborating RSP projects, including the definition of known relationships. Based on this, an initial analysis of the generated data will be conducted.

This will be followed by continuous correction and development of the analysis system in collaboration with the RSP projects.