A method for generating finite element models of wood boards from X-ray computed tomography scans. (February 2022)

• Record Type: Journal Article
• Authors: Huber, Johannes A.J.; Broman, Olof; Ekevad, Mats; Oja, Johan; Hansson, Lars
• Is Part Of: Computers & structures. Volume 260(2022)
• Publication Date: 2022-02
• Subjects: CoV coefficient of variation -- CT computed tomography -- FCS fibre coordinate system -- GST gradient structure tensor -- HU Hounsfield unit -- MoE modulus of elasticity -- MPD minimum pith distance -- R2 coefficient of determination -- RMSE root mean squared error -- SD standard deviation; Fibre reconstruction -- Feature reconstruction -- CT scanning -- Strength grading -- Image analysis -- Gradient structure tensor; Structural engineering -- Data processing -- Periodicals; Electronic data processing -- Structures, Theory of -- Periodicals; 624.171
• DOI: 10.1016/j.compstruc.2021.106702
• Languages: English
• ISSNs: 0045-7949

Abstract

Highlights: Reconstruction of board geometry, pith and knots from computed tomography scans. New fibre reconstruction algorithm based on image analysis of CT density data. 3D and 1D finite element models to simulate four-point bending test. Estimation of local and global bending stiffness (MoE) and bending strength. Coefficient of determination ( R 2 ⩽ 0.83 ) for bending strength and ( R 2 ⩽ 0.93 ). Abstract: A method is presented for reconstructing the geometry, the pith, the knots and the local fibre orientations in timber boards, based on X-ray computed tomography scans. The local fibre deviations around knots were found by a new algorithm, based on image analysis. The experimental data comprised tomography scans, eigenfrequency measurements and four-point bending tests of 20 Norway spruce boards. 3D and 1D finite element models of the pure bending zone of the bending tests were created, accounting for the exact board geometry and the reconstructed fibre deviations. A purely density based, a purely eigenfrequency based, and a mixed constitutive law were compared. Model estimations showed a high coefficient of determination ( R 2 ) for global modulus of elasticity (MoE) ( R 2 ⩽ 0.93 ), local MoE ( R 2 ⩽ 0.87 ), bending strength ( R 2 ⩽ 0.83 ), and the location of initial failure. Constitutive laws accounting for eigenfrequency showed the most accurate results. In the future, adapting the method for logs could enable analyses of boards before sawing.