Metamodel assisted optimization of glued laminated timber beams by using metaheuristic algorithms

Abstract An efficient use of the raw material in glued laminated timber (GLT) beams is commonly achieved by inserting lamellas of lower quality in less stressed areas, usually in the middle of a beam. But, this rather simple method leaves room for improvement. In particular, the morphology of a board and its location in the beam setup is significant, since only this information and the actual loading situation allows a proper evaluation of weaknesses. Therefore, a new optimization method was developed, able to take mechanical property distributions as well as the occurring stress states within each wooden board into account. Subsequent to an automatic knot reconstruction and determination of effective local stiffness distributions of all boards, the beams are analyzed using a finite element (FE) model. This information is further exploited to find optimal beam setups out of a sample of boards. However, as the complexity of this optimization task quickly increases with the number of boards, metaheuristic optimization algorithms were developed. Additionally, the evaluation of the computationally expensive FE model is bypassed by a metamodel, capable of approximating the desired performance parameter of any beam. Comparing the various optimization approaches to common GLT beam production methods, maximum deflection can be reduced by 15%–20%.