A novel artificial intelligence-aided system to mine historical high-performance concrete data for optimizing mixture design

Designing proper HPC mixtures is an onerous task, as many different chemical substances are used to produce HPC of differing specifications and performance characteristics. In this study, a novel, artificial-intelligence-aided system is developed to effectively minimize the cost of producing HPC of variable compressive strength, workability, and durability requirements. In the first phase, a symbiotic organism search-optimized least squares support vector machine (SOS-LSSVM) is developed using historical data to predict HPC properties to verify that these properties conform with requirements. In experimental testing conducted for this study, the SOS-LSSVRM achieved MAPE values of 17.44% 16.31%, and 22.70%, respectively, for predicting compressive strength, workability, and durability. In the second phase, symbiotic organism search (SOS) is used to identify the lowest-cost HPC mixture that attains the specified mechanical requirements. The results demonstrate that the proposed artificial intelligence-aided system is effective in guiding the economical design of HPC mixtures that comply with ACI and DMDA standards.