Genetic algorithm optimisation of heavy timber trusses with dowel joints according to Eurocode 5

Abstract

This paper presents the optimisation by genetic algorithms of the members and joints of heavy timber trusses fastened by mechanical joints with dowels and metal plates. This is achieved by programming in Matlab twenty-four elitist genetic algorithm models which, in combining different variables, provide an estimate of the geometry of the truss, the cross-section dimensions, and the number of dowels required. Trusses with various numbers of nodes are evaluated, with 6, 10, and 14 segments in the top chord. After validation of the results with a preliminary bibliography, an analysis is made of the effect of slipping of the joints in accordance with current standards, studying how such slipping affects the dimensioning of the trusses. Subsequently, to achieve the greatest economy in their design, the depth of the trusses is optimised. Genetic algorithms constitute an effective optimisation tool for the design of heavy timber trusses. They led to reduced costs compared with previous techniques described in the literature. In each type of truss studied, the optimal volume of timber was very similar for all cases, regardless of the number of upper chord members or joint slip. The cost of the truss therefore depended strongly on the number of joints, with the most economical solution being that with the fewest joints. Joint slip was important when the serviceability limit state was determinant, with its inclusion increasing the cost.