Structural optimization of self-supported dome roof frames under gust wind loads

Abstract

Dome roofs are large structures often subject to variable wind, snow and other loading conditions, in addition to their own weight. A wide variety of structural designs are used in practice, and finding the optimal arrangement of trusses or girders, along with suitable section properties, is a common subject for structural optimization studies. This thesis focuses on self-supported dome roofs for fuel storage tanks, and a variety of optimization techniques are adapted, developed and compared. Various load conditions have been compared using detailed fluid and stress analysis in ANSYS. From results for full and empty storage tanks, with wind and/or snow external loads, the worst cases are for wind loading alone, i.e., snow loading counters the lift force from the wind. Consequently, the case of an empty fuel storage tank subject to wind loading is used as the basis for the structural optimization. To speed up the optimization, a simplified frame analysis was developed in Matlab and integrated with the optimization code. In addition, the wind loads were modelled in ANSYS for a range of dome radii and imported into the Matlab, and a number of different dome designs were used as case studies: these were ribbed, Schwedler, Lamella and geodesic. The principal method used to optimize the frame is Morphing Evolutionary Structural Optimization (MESO), in which an initial overdesigned frame is iteratively analysed and reduced in overall weight by reducing the sections of key frame members. The frame is progressively weakened, but without compromising the structural integrity, until it is no longer possible to reduce the weight. However, there are additional parameters that MESO is not suited to, such as dome radius and those affecting the overall structure of the dome frame (numbers and placements of rings, etc.), and a variety of metaheuristic optimization techniques have been studied: Artificial Bee Colony (ABC), Bees Algorithm (BA), Differential Evolution (DE), Particle Swarm Optimization (PSO) and Simulated Annealing (SA). These can be used instead of MESO, or in a hybrid form where MESO optimizes the frame member sections. Although the focus in this thesis is on minimizing the total structural weight, the importance of other characteristics of the design, especially structural stiffness, is considered and also integrated with the MESO process. The hybrid methods MESO-ABC and MESO-DE performed best overall.