Automatic structural optimization design using sensitivity-based method

Abstract

Traditional structural design involves repeated modeling and parameter tuning of buildings, which is time-consuming work. To solve the problem, the automatic structural optimization design of a multi-lateral force resist system is proposed out. The aim is to minimize the cost of structural materials based on initial determinate structural model subjected to multiple constraints according to the design codes. A simplified model including mega braces, mega frames, truss systems, shear walls etc. is employed, which can be easily generalized to other types of structures. The finite element method is used to obtain the structural responses by ETABS, while two modified algorithms are adopted for structural optimization. The first one is total penalty cost-based constant incremental sensitivity analysis method that is introduced with full-level constraints as a whole and the adaptive adjustment mechanism. The second one is structural constraint-based Hooke-Jeeves algorithm considering different building modulus and discreteness of design variables. The influence of their parameter settings on optimization performances are discussed such as different initial designs. The results show that the sensitivity-based methods are useful tools for automatic optimization design of an existing building with known structural parameters and the recommended parameter settings of algorithms are provided. As a result, both the structural safety and economic benefit of the discussed structure can be achieved with high efficiency. A prompt convergence rate and a good approximation to a deterministic solution can also be observed. It can provide a reference to the structural design in practice.