Machine strength grading is a production process in the sawmill industry used to grade sawn timber boards into different strength classes with specific characteristic values of the bending strength, modulus of elasticity (MOE) and density. These properties are called grade determining properties. Each of these is predicted on the basis of a statistical relationship between the property and a so-called indicating property (IP), which is based on non-destructively assessed board properties. In most cases, the prediction of strength is crucial for the grading. The majority of commercial grading machines rely on a statistical relationship of strength to an IP, which is either a global dynamic MOE or an averaged flatwise bending MOE measured over a board length of about one meter. The problem of today’s machine strength grading is that the accuracy of the strength prediction is rather poor with a coefficient of determination of about R2 ≈ 0.5 − 0.6. One consequence of this is that much of the strength potential of timber is unused.

The intention of this research is to contribute to a long-term goal, which is development of a method for prediction of bending strength that is more accurate than the methods available today. The research relies on three hypotheses. First, accurate prediction of bending strength can be achieved using an IP that is a localized MOE value (determined over a short length) that represents the lowest local bending stiffness of a board. Second, knowledge of the local bending stiffness with high resolution along a board’s longitudinal direction can be established on the basis of fibre direction within the board in combination with dynamic MOE. Third, fibre directions in the interior of a board can be determined by application of fibre angle models utilizing data of fibre directions on the board’s surfaces obtained from tracheid effect scanning. Following these hypotheses, this work has included laboratory investigations of local material directions, and development of models for fibre directions of the interior of boards. The work also included application of one-dimensional (1D) analytical models and three-dimensional (3D) finite element models of individual boards for the mechanical behaviour, analysis of mechanical response of boards based on experiments and based on the suggested models. Lastly, the suggested models were evaluated by comparisons of calculated and experimentally determined local bending stiffness along boards, and of predicted and experimentally determined bending strength.

The research contributes with in-depth knowledge on local fibre directions close to knots, and detailed information on variation of the local bending stiffness in boards. Moreover, fibre angle models for fibre directions in the interior of boards are presented. By application of the fibre angle models in the 3D model of the whole board, the local bending stiffness along timber boards can be determined over a very short length (l < 50 mm). A comparison with results determined on an experimental basis show a very close similarity implying that the applied models are sufficient to capture the variation of local bending stiffness, caused by knots and fibre distortions, with very high accuracy. Furthermore, it is found that by means of IPs derived using the suggested models, bending strength can be predicted with high accuracy. For a timber sample comprising 402 boards, such IPs results in coefficient of determination as high as R2 = 0.73. However, using IPs based on the 3D finite element model did not improve the R2 value achieved when using the IPs based on the 1D model.

Machine strength grading of sawn timber is a sawmill process in which density, modulus of elasticity (MOE) and bending or tensile strength are predicted such that the timber can be assigned to strength classes. The predictions of these properties are performed using one or several so-called indicating properties (IPs), which represent a board property, or combination of board properties, measured non-destructively. A limitation of today’s strength grading is that the IPs applied in the industry for prediction of strength, in general, are based on rather weak statistical relationships between IPs and strength properties, which in turn results in poor material utilisation. It is well known that the strength of sawn timber is associated with the presence of knots and their surrounding fibre disorientations. Local fibre direction at surfaces of softwood can be determined by means of the light scattering that occur when a wood surface is illuminated by a dot-laser, i.e. by application of the so-called tracheid effect. Lately, IPs based on such measurements have been developed, and some of the suggested IPs have a strong statistical relationship to bending strength. The purposes of the research presented in this thesis are to contribute with knowledge of possibilities and limitations of the tracheid effect and of data of fibre directions in the vicinity of knots, to evaluate if information of fibre directions at surfaces of Norway spruce sawn timber can be used to achieve a better material utilisation of glulam lamellae and finger-jointed timber, and to provide insight regarding the grading regulations in Europe. Results presented herein show that knots and fibre direction within the interior of boards can be modelled on the basis of data obtained by means of the tracheid effect, but also that a previously proposed method to determine out-of-plane fibre angles gives poor accuracy. As regards grading of glulam lamellae, an IP based on fibre directions and dynamic MOE is proposed for prediction of tensile strength. The latter is used when grading glulam lamellae. Application of the proposed IP resulted in substantially increased yield in strength classes. It is also shown that this IP is applicable for boards with sawn as well as with planed surface finish. Regarding current regulations for machine strength grading in Europe, results indicate that grading based on global board properties give higher yield than what is appropriate.