Prediction And Optimization Of Bearing Capacity Of CFST Columns Using Artificial Intelligence Techniques

A composite structure under confinement can become stronger.A concrete-filled steel tube(CFST)column is a composite structure in which the concrete is confined by a steel tube.The circular concrete-filled steel tube(CCFST)column is a commonly used composite structure of these types.The CCFST column is the most popular and applicable type because of its round cross-section and superlative structural performance.The CCFST column provides an appropriate confining pressure for the concrete compared to other types and enhances the interaction between the steel tube and the concrete and provides a larger load-bearing capacity for the column.The main objectives of this study are:i)to use gene expression programming(GEP)(the newest artificial intelligence technique)to analyse data and generate an equation to predict the bearing capacity of concrete-filled steel tube columns and to compare the results of GEP with ANN and PSO-ANN to demonstrate the capability of GEP;ii)to optimize the bearing capacity of CFST columns using artificial bee colony(ABC),whale optimization algorithm(WOA),and invasive weed optimization(IWO)techniques;and iii)to assess the capability of GEP by comparing its results with a finite element analysis(FEA)of the CFST columns.Based on the results,the following conclusions were obtained.(1)Proposed GEP equations to effectively predict the bearing capacity of CFST composite columns and shows higher performance than that of PSO-ANN.This study adopted several artificial intelligence(AI)techniques to predict the bearing capacity of CFST columns in two phases:prediction and optimization.In the prediction phase,the bearing capacity values of CFST columns were estimated using the proposed GEP.With 303 datasets,the outer diameter,concrete compressive strength,tensile yield stress of the steel column,steel cover thickness,and length of the used samples were considered as model inputs.After a series of analyses,five GEP equations were developed,and the best one was determined based on the coefficient of determination(R~2).To validate the reliability of the GEP prediction model,both ANN and PSO-ANN models were implemented for verification and comparison.The outer diameter,concrete compressive strength,tensile yield stress of the steel column,thickness of the steel cover,and length of the samples were model inputs.After a series of analyses,the best predictive models were selected based on the R~2 results.In the GEP-based tree equation,R~2 values of 0.928 and0.939 were used for the training and testing datasets,respectively.The PSO-ANN model used R~2 values of 0.910 and 0.904,and the ANN model used R~2 values of 0.895 and 0.881for the training and testing datasets.The results demonstrated that the GEP provides a higher bearing-capacity performance than that of the PSO-ANN and ANN models.(2)Optimized ultimate bearing capacity of CFST columns.In the optimization phase,an ABC was developed to maximize the results of bearing capacity.As a result,the ABC was able to design input parameters such that the maximum values of bearing capacity were obtained.This result indicates that the ABC is a powerful optimization algorithm for solving these kinds of engineering problems.To obtain better results,the WOA and IWO were also adopted and their results were compared with those of the ABC.The WOA was selected and developed to maximize the results of bearing capacity.Based on the obtained WOA results,the bearing capacity of CFST columns was significantly maximised by increasing the bearing capacity to 23436.63 k N.The IWO technique also maximised the bearing capacity by considering the selected model.From a comparison of the IWO and ABC techniques,it was found that,for both optimization algorithms,input parameters can be designed to obtain the maximum bearing capacity.The bearing capacity of the CFST columns using the ABC and IWO techniques indicates that the IWO has a better capability of maximizing the bearing capacity.Thus,the IWO technique can optimize similar problems with a high rate of performance.(3)GEP provides better results than those of FEA.To demonstrate the power of AI techniques compared to FEA methods,32 samples were modelled,and their ultimate bearing capacities were obtained and compared with the GEP equation output results as well as the experimental results.The comparison results showed that the values obtained from GEP were closer to the real experimental data than were the values obtained from FEA by up to 12%.This demonstrates the accuracy and power of the prediction equation obtained from GEP for these types of CFST columns.