An Innovative Multi-Objective Optimization Approach for Compact Concrete-Filled Steel Tubular (CFST) Column Design Utilizing Lightweight High-Strength Concrete

Abstract

Incorporating sustainability principles into the optimization of Concrete-filled steel tubular (CFST) columns can lead to reduced material consumption, lower costs, and decreased emissions, aligning with the goals of energy efficiency and sustainability in construction and building materials. However, incongruities exist within international standards regarding the computation of the ultimate load capacity for compact CFST columns subjected to eccentric loading, especially when employing lightweight high-strength concrete. In order to address these discrepancies, an exhaustive dataset comprising compact CFST columns of both rectangular and circular profiles (including built-up and hot-rolled configurations) was systematically compiled from published academic literature. To evaluate the precision of prevailing design codes in estimating the strength of compact concrete-filled steel tubular (CFST) sections under eccentric loading, the experimental findings were juxtaposed with estimates derived from AISC 360-16 and Eurocode 4. Following this, an all-encompassing three-dimensional finite-element model was constructed. This model was designed to forecast the performance and strength characteristics of compact CFST columns subjected to eccentric loading, as well as to investigate the essential axial force-moment (P-M) interaction behavior with respect to the material strength ratio ([[EQUATION]]/[[EQUATION]]). In the second phase of the study, an artificial neural network (ANN) model with a backpropagation algorithm was developed to estimate the axial load capacity of circular and rectangular eccentrically loaded CFST columns. Furthermore, by incorporating input parameters such as axial load, eccentricity in the x and y directions, column length, and other relevant factors, a multi-objective optimization methodology was devised to determine the optimal geometry for CFST columns comprehensively. The results revealed that Eurocode 4 provided better predictions of the experimental axial capacity compared to the theoretical value, [[EQUATION]]/[[EQUATION]], with a lower scatter histogram than AISC 360-16. The mean and standard deviation for Eurocode 4 were estimated at 1.07 and 0.22, respectively, compared to 1.21 and 0.29 for AISC 360-16. Several statistical metrics were employed to verify the efficiency of the proposed ANN model against AISC 360-16 and Eurocode 4. The novel information model and optimization method demonstrated remarkable precision, especially in scenarios involving high-strength concrete. This holds the potential to streamline generative design processes in forthcoming computational intelligence-based structural design platforms.

Keywords: Reinforced Concrete, Steel tube, CFST column, Composite, Finite Element Modeling, Artificial Intelligence, Multi-Objective Optimization, Design code