Paper quality improvements and production cost reduction by using new Phys-AI technology (PAIT)

Important results from the project

The project provided valuable insights despite not fully achieving its initial goals. Originally aimed at using machine learning to assess paper quality at the dryer section’s end, the focus shifted to paper breakage in the winding section. This complex issue, with multiple root causes and limited prior knowledge, was explored through sensor installations to enhance datasets. The study improved dataset understanding, aiding the paper industry´s digital transition and future quality assessments.

Expected long term effects

The project aims to develop digital twins for the paper industry, enabling a seamless shift to Industry 5.0. These virtual replicas will support real-time monitoring, predictive maintenance, and optimization through AI, IoT, and data analytics, enhancing efficiency. Additionally, it focuses on improving dataset standardization for Supervised Machine Learning, ensuring high-quality data collection, labeling, and processing to enhance AI model accuracy and automation in industrial operations.

Approach and implementation

The project was split in 5 work packages: coordination and management, investigation and scale of the problem, data preparation and analysis, mechanical tests and interpretation of results. To ensure collaboration, regular online meetings with all partners have been organized. The datasets (Holmen/SPM) and test samples have been shared through cloud services and send by mail, where each partner performed specific analysis and communicate their results in the meetings and deliverables.