A hybrid gradient-based/metaheuristic method for Eurocode-compliant size, shape and topology optimization of steel structures. (15th July 2021)

• Record Type: Journal Article
• Authors: Dillen, Wouter; Lombaert, Geert; Schevenels, Mattias
• Is Part Of: Engineering structures. Volume 239(2021)
• Publication Date: 2021-07-15
• Subjects: Structural optimization -- Eurocode design -- Real-life optimization -- Genetic algorithm -- Sequential quadratic programming -- Active set; Structural engineering -- Periodicals; Structural analysis (Engineering) -- Periodicals; Construction, Technique de la -- Périodiques; Génie parasismique -- Périodiques; Pression du vent -- Périodiques; Earthquake engineering; Structural engineering; Wind-pressure; Periodicals; 624.105
• DOI: 10.1016/j.engstruct.2021.112137
• Languages: English
• ISSNs: 0141-0296

Abstract

Highlights: Simultaneous size, shape and topology optimization involves mixed design variables. Metaheuristic methods can handle discrete design variables. Gradient-based methods are efficient for continuous design variables. Both methods are combined in a new hybrid optimization strategy. Benchmark problems take into account all relevant Eurocode constraints. Abstract: The production and processing of building materials is responsible for a significant share of global greenhouse gas emissions. Nevertheless, in current building practice, a lot of material is wasted, because load-bearing structures are often grossly overdimensioned. Numerical optimization tools have the potential to reduce the consumption of structural materials, but the lack of customized algorithms prohibits their use in daily design practice. This is because real-life structural optimization problems are complex, involving both discrete and continuous design variables as well as large numbers of building code constraints. In the literature, such problems are usually solved using metaheuristic (often genetic) algorithms, although gradient-based algorithms are better suited for continuous design variables. In this article, a new hybrid gradient-based/metaheuristic algorithm is proposed. It combines gradient-based and metaheuristic methods in a nested approach, exploiting the strengths of both: the ability to handle discrete design variables and fast convergence in terms of continuous design variables. The applicability and usefulness of the new hybrid algorithm is demonstrated in two realistic case studies to minimize the weight of the structure, taking into account all relevant design rules of Eurocode 3. Within identical computing time, the hybrid algorithm obtains significant material savings compared to a conventional genetic algorithm.