Prediction of tensile strength of concrete using the machine learning methods

Machine learning methods have been utilized to simulate and forecast the tensile strength of concrete. Initially, various deep neural networks were simulated using different optimizers to change the number of neurons present in every hidden layer. The models take 4 features as input and the result will be predicted by the trained artificial neural network models (ANN) models. These models take the content of water, cement, sand, and gravel. As input, the comparisons that have been done between the predictions of the models and experimental results have shown that the models make predictions about the tensile strength of concrete with an accuracy level of approximately 80 percent. The data used are the result of 113 concrete tensile strength tests. Next, a comparison was made between the ANN-based formula and traditional machine learning models including Counter Propagation Neural Nets (CPNN), Radial Basis Function Neural Network (RBFNN), K-Nearest Neighbors (KNN), XGboost, Cat boost (CB), Decision Tree (DT), Random Forest (RF), Extra Trees (ET), Light GBM, Ada Boost (AB), Bagging (BA), Gaussian Process (GP), Support Vector Machine (SVM), and Ensemble methods to show the accuracy and low loss of the ANN model. The findings suggested that the artificial neural network (ANN) model produced more precise forecasts (R 2 = 0.9397) compared to conventional machine learning models. Based on the sensitivity analysis results obtained using the ANN model, it can be inferred that the cement parameter is the most significant feature for accurately estimating the tensile strength of concrete using this particular dataset. © 2023, The Author(s), under exclusive licence to Springer Nature Switzerland AG.