An Estimation Proposal for Engineering Properties of Modified Concrete by using Standalone and Hybrid GRELM

The presented study pertains to an attempt to propose a novel prediction model to predict the flexural and compressive strengths of concrete modulated using steel fiber (SFb) and silica fume (SF). A completed experimental investigation is adopted for the current study, and a research plan is employed. Three different superplasticizers amount (SP), different SF replacement ratios, and a constant amount of SFb were used by the weight of cement to meet the C25 target strength. A sum of 16 distinct mixtures designed by changing SP and SF ratios were developed. Furthermore, SFb was added at a fixed rate of 65 kg/m3 to all planned concrete mixes. In addition, SFb was used to create a 16-mix design. Finally, a total of 32 distinct mix designs were created. Produced, hardened specimens were exposed to two different curing conditions. This research uses the mechanical characteristics of concrete treated with SF, SP, and SFb to estimate by conducting standalone and hybridized generalized extreme learning machine (GRELM) algorithms based on available experimental data in terms of the metaheuristic aspects of this work. With continuous input data, four separate data sets were constructed. Compressive strength and flexural strength were estimated separately. With the aid of the Grey Wolf Optimization (GWO) and Particle Swarm Optimization (PSO) algorithms, binary and ternary hybrid approaches were developed and tested on the data. Four distinct estimation models were suggested. Two quality metrics were used to evaluate the estimation performance: Root Mean Square Error (RMSE) and correlation of determination (R2). The estimation results showed that the hybridized GRELM-PSO-GWO estimation model that was built for prediction was relatively successful in all sets.