Employing the optimization algorithms with machine learning framework to estimate the compressive strength of ultra-high-performance concrete (UHPC)

Ultra-high-performance concrete (UHPC), with the highest resistance capacity against axial loads, is composed of various ingredients compared to other typical concretes, including fly ash and silica fume; eco-friendly materials also have an economical price. Such high-resistant constructional materials are seen as a suitable aggregate to use in most practical projects. The compressive strength (CS) of concrete, as one of the important variables in engineering fields, can be estimated using smart approaches based on ingredients as inputs fed to the mathematical model. Consequently, the current study modeled the CS values using a machine learning technique of Support vector regression (SVR) accompanied by the grasshopper optimization algorithm (GOA) and arithmetic optimization algorithm (AOA), tunning the SVR to appraise the CS accurately. In developing AOA-SVR and GOA-SVR frameworks, several metrics were used to assess the ability and accuracy of models. As a result of the present study, it can be concluded that the machine learning method in hybrid form has the ability to predict for saving time and energy. In general, in comparing the results of hybrid models, AOA had the most suitable combination with SVR compared to GOA, with R-2 = 0.901 and RMSE = 9.986 in the training phase. In addition, AOA-SVR improved its performance by R-2 = 0.917 and RMSE = 9.525 in the test phase.