Theses

Master Theses

Estimating the tensile strength of beech wood based on image analyses and Machine Learning

Desciption

Fibre direction data computed from images of European beech (Fagus sylvatica L.) boards without obvious visual defects has been successfully used to improve the prediction of their tensile strength parallel to the grain. Machine-learning (ML) methods applied to the same data-set have also shown a good prediction accuracy, but the total number of boards in the data-set was less than 50 and this limited the potential of ML methods. The objective of the proposed master thesis is to create a larger data-set of boards and extend the use of ML methods to estimate the tensile strength parallel to the grain of beech wood boards. The work will consist of: i) documenting European beech boards (ρ, E_dyn, and images); ii) testing the boards (f_{t,0,experimental}); iii) further developing the ML models to estimate the tensile strength (f_{t,0,estimated}) based on the raw images of the boards.

Supervision

Mark Schubert (Empa); Pedro Palma (Empa); Ingo Burgert (ETH Zürich)

Additional info

Thesis to be conducted at Empa, Dübendorf.