The main purpose of the pursued research presented in this project was to study whether a two-dimensional finite element model of thin timber members containing a single live knot and integrating surface grain data gathered by a laser scanner, could predict the load bearing capacity of such members with a satisfactory accuracy. In the modeling process, it was hypothesized that the scanning data could help obtain some additional information regarding the inner structure of the members in order to improve the predictions.
Furthermore, it was put forward that the predictions could be correctly determined by the use of fracture mechanics theory. An analysis of the data files provided by the scanner demonstrated that the grain data is meaningless over the knots and therefore that no well defined boundary exists in the numerical grain pattern between the knots and the surrounding clear wood. Nevertheless, it was noticed that other data could help define the centre of the knots and might correspond, at least coarsely, to the orientation of the fibres in the out-of-plane direction (relating to the surface of the members).
The data was then implemented into the numerical model both for attributing some particular mechanical properties to the knots in comparison to the clear wood and for attributing an out-of-plane angle (or dive angle) to the fibres located in the knots and their vicinity. The model was simulated in bending in order to determine the strains and the stress concentrations of two selected boards. The simulations were carried out with various configurations relating to the mechanical properties of the knots and to the dive angle of the fibres. Meanwhile, the boards were tested to failure in a bending test rig and the system Aramis® measured the strains and took photos during loading and up to the instant of failure. Afterwards, the numerical results were compared to the experimental tests to determine how the dive angle and the material properties can make the model tally with reality.
The prediction of the load bearing capacity was achieved thoroughly for one board but resulted, at the best, in a 31.7% underestimation with respect to the actual strength of the board. Several additional predictions were made with modifications of the material properties (because most of the values came from the literature and not from measurements) and emphasized their influence on the variations of the predictions.
Finally, the closest prediction was 14.6% lower than the actual strength. From the analysis of the model configurations and the predictions, it was concluded that the small number of tests does not enable one to distinguish the part of the material properties from the part of the modeling approach and from the part of the accuracy of the measurements in the gap between the predictions of the load bearing capacity and the actual strength of the board. However, some important questions regarding the modeling were raised through the analysis and the conclusions and some prospects were proposed for further research.
Source: Linnaeus University
Author: Luca, Matthieu