Master's thesis; Machine learning for cellulose fibres

Master's thesis; Machine learning (ML) models that can estimate the percentage of each fibre type contribution in a paper boardBackgroundThe thesis forms part of the VINNOVA financed project “AI-Tomo: Accelerated materials characterisation by AI and X-ray tomography” (https://www.vinnova.se/en/p/ai-tomo-accelerated-materials-characterisation-by-ai-and-x-ray-tomography/). The goal of AI-TOMO is to develop AI algorithms for fast, effective segmentation and quantification of 3D and 4D X-ray tomography data to accelerate materials development. AI-TOMO is a close collaboration between the research providers RISE and Lund University, the synchrotron facility MAX IV, and the companies Billerud and TetraPak. This master thesis will be a cooperation between material scientists at Billerud and AI researchers at RISE with support from X-ray tomography experts at Lund University.Billerud aims to create new materials that can replace fossil-based materials with more sustainable alternatives based on forest raw materials. The major challenge for cellulosic materials is to improve certain properties, such as extensibility, strength or barriers to water vapor and other gases and to produce these materials in a resource- and energy-efficient way. Billerud produces paper boards with fibres from different trees which have different properties and so different combinations of fibre types produce paper boards with different properties. It is therefore key to monitor the composition of paper boards to ensure they are formed by the desired fibre combination. Currently, this is carried out by experienced microscopists who can visually tell if the composition of the paper board is correct. The aim of this master thesis project is to develop ML models that can streamline the process by automatically identifying the contribution of each paper fibre type in the board.DescriptionTo develop machine learning (ML) methods that can estimate the percentage of each fibre type contribution in a paper board. This task can be defined as a classification problem where the goal is to estimate the presence of each class in a given image, taking different fibre types as different classes and X-ray tomography as input image.Key ResponsibilitiesLab paper sheets are available with known mixtures of different fibers, where distribution of fiber dimension has been measured a priori. The 3D structures of these samples will be recorded using X-ray tomography, which is performed in collaboration with Prof. Stephen Hall at Lund University (LTH Faculty), which is part of the AI-TOMO project. Main task. Develop an AI model that can estimate the percentage of fibre types in the X-ray tomography image. To complete the work, the following activities need to be performed:Pre-process an existing labelled dataset of X-Ray tomography images of paper boards with the corresponding fibre percentage compositions and render it AI-ready.Design and develop an AI system for image classification Train and validate model with the gathered datasetEvaluate results and test for robustnessWriting of MSc thesis and reporting at Billerud, the university there the student is enrolled and the AI-TOMO-project.QualificationsTo be able to successfully contribute to this degree project, we believe that you:Are a master student in computer science, mathematics, applied physics with machine learning orientation or similar.Have experience with machine learning Have knowledge in materials science, X-ray imaging and tomography Have solid oral and written English skills.TermsRecruiting manager: Sepideh Pashami, PhD (sepideh.pashami@ri.se)Scientific supervision will be given by: Delia Fano Yela, delia.fano.yela@ri.se; Smita Chakraborty, smita.chakraborty@ri.se ; Eoin Walsh eoin.walsh@solid.lth.se Industrial supervisor: Mikael Nygårds, Mikael.Nygards@billerud.com Examiner needs to be contacted by master student at the universityLocation: Office locations in Lund (RISE) or Stockholm (Billerud) can be arrangedApplication deadline: November 14th 2025Timeline: January -June 2026. A detailed schedule will be defined together with the candidate. The expected time is 20 weeks full time (40 hours/week) work, roughlyCredits: 30 HPCompensation: 30000 SEK upon a successful completion of a high-quality thesis.For any questions, please contact Delia Fano Yela, delia.fano.yela@ri.se; Smita Chakraborty, smita.chakraborty@ri.se; or Mikael Nygårds, Mikael.Nygards@billerud.com.Welcome with your application!