Hybridization of metaheuristic algorithms with adaptive neuro-fuzzy inference system to predict load-slip behavior of angle shear connectors at elevated temperatures

Steel-Concrete Composite floor systems are one of the essential components in the construction industry. Recent studies have shown that fire-induced problems damage shear connectors and change the behavior of composite systems. To predict the performance of connectors, experimental tests are generally conducted at elevated temperatures or fire conditions. However, these tests need plenty of time, cost, and effort. This paper aims to propose a soft computing (SC) approach to predict the behavior of angle shear connectors at elevated temperatures. For this purpose, an adaptive neuro-fuzzy inference system (ANFIS), a particle swarm optimization (PSO) algorithm, and a genetic algorithm (GA) are hybridized and a novel ANFIS-PSO-GA model is proposed. To evaluate the performance of the ANFIS-PSO-GA model, a radial basis function network (RBFN) along with an extreme learning machine (ELM) are also developed. Finally, the performance of the ANFIS-PSO-GA, RBFN, and ELM are compared in the terms of different statistical indicators. The results of the paper show that the SC approach is applicable in the behavior prediction of angle connectors at elevated temperatures. Besides, it was concluded that the ANFIS-PSO-GA model can provide better estimations of load and slip in comparison with those of RBFN and ELM models.

Automated summary

Steel-Concrete Composite floor systems are crucial in construction, but fire-induced issues can damage connectors and alter system behavior. This paper introduces a soft computing approach to predict connector behavior at elevated temperatures, demonstrating superior performance with the ANFIS-PSO-GA model.