Samples of European beech (Fagus sylvatica) were used for this study. Logs of these samples covered a scatter of mild-to-strong curvatures and the boards of these samples covered strong fiber deviations. This study consists of two separate parts: (1) log reconstruction and optimization of the cutting pattern, and (2) board reconstruction and strength prediction. Information about the internal quality of the logs is missing in this study, as laser scanning has been used for surface reconstruction of logs. Therefore, two separate steps were implemented here. (1) Influence of cutting pattern and board-dimensions on yield were analyzed. For this step, 50 logs were checked. (2) A more advanced numerical method based on the finite element (FE) analysis was developed to improve the accuracy of tensile strength predictions. This step was performed, because visual grading parameters were relatively weak predictors for tensile strength of these samples. In total, 200 beech boards were analyzed in this step. However, due to the geometrical configuration of some knots, the reconstruction and numerical strength prediction of 194 boards out of 200 boards were possible. By performing tensile tests numerically, stress concentration factors (SCFs) were derived, considering the average and maximum stresses around the imperfections. SCFs in combination with the longitudinal stress wave velocity were the numerical identifying parameters (IPs), used in the nonlinear regression model for tensile strength prediction. The influence of the combination of different numerical parameters in the developed non-linear model on improving the quality of the strength prediction was analyzed. For this reason, improvement of coefficient of determination (R2) after adding each parameter to the multiple regression analysis was checked. Performance of the developed numerical method was compared to the typical grading approaches [using knottiness and the dynamic MoE (MoEdyn)], and it was shown that the coefficient of determination is higher, when using the virtual methods for tensile strength predictions.