Prediction of the compressive strength of strain‐hardening cement‐based composites using soft computing models

Abstract

Strain‐hardening cement‐based composites or engineered cementitious composites (ECC) is concrete produced using randomly distributed short polymer fibers. It is very ductile compared to conventional concrete. Compressive strength (CS) is a critical property used as a quality control tool to evaluate the strength of concrete implemented in the structural provisions and mix designs. Accordingly, to save cost and time for testing, it is essential to provide a predictive model to forecast the CS of the concrete mixtures using machine learning and modeling techniques. In this study, different modeling tools are used to propose analytical models to predict the CS of ECC mixtures, such as linear regression (LR), multi‐expression programming (MEP), artificial neural network (ANN), and Gaussian process regression (GPR). A total of 210 data were collected from the literature and used to train and test the developed model. The fly ash‐to‐cement ratio ranged from 0 to 5.6, water to binder ratio from 0.19 to 0.56, different superplasticizer and fiber content, and curing times of 1 to 180 days. Based on the evaluation of the developed models, the ANN model is superior to other developed models with a high coefficient of determination (R2), root mean squared error (RMSE), mean absolute error (MAE), and scatter index (SI). The sensitivity analysis of the input parameters' effect on the CS prediction indicates that the curing time and the fly ash‐to‐cement ratio are essential in forecasting ECC's CS.

Artificial rabbit optimization‐based ANFIS model development for predicting the compressive strength of GGBFS‐based concrete

“…To determine how the rule will affect every node's contribution to the result, the nodal function is used. Equation (25) shows the result that is produced when the linear input signals and the normalized signals from previous nodes are added together.…”

“…Based on the evaluation of the developed models, the ANN model was superior to other developed models with a proper metrics. 25 Using soft computing methods could be of great interest in predicting the f c of ultra-high-performance fiber reinforced concrete (UHPFRC). The results showed a high prediction accuracy of the f c of UHPFRC using ANN models.…”

Strain rate sensitivity of strain‐hardening fiber‐reinforced concrete subjected to dynamic direct tensile loading

“…20 ML shows high effectiveness and accurate performance in predicting the complex relationship among input and output variables, thus it is expected to offer an apparent benefit over traditional regression methods. 21 Although several recent studies have proposed ML models to predict the tensile resistance of SHFRCs at a static rate, 22,23 very few studies have focused on forecasting the strain rate sensitivity of SHFRCs in tension using ML. From the literature review, only Yang et al 22 proposed the use of ML in predicting the strain rate sensitivity of tensile strength of SHFRCs.…”