Development of compressive strength prediction platform for concrete materials based on machine learning techniques

Abstract

With the continuous development of artificial intelligence, machine learning (ML), as an important branch, is used to promote the digitalization of concrete. Considering that the verification of the strength corresponding to the concrete mixture requires strict curing system and cycle, high cost and low efficiency, ML technology is employed to predict the 28-day compressive strength of ordinary concrete for the support of engineering practices. The aim is to develop an efficient platform for predicting the compressive strength of concrete materials. Ridge Regression (RR), Least Absolute Shrinkage and Selection Operator (Lasso), Elastic Net Regression (EnR), Support Vector Regression (SVR) and K-Nearest Neighbor (KNN) are established, of which efficiency and accuracy are analyzed and calculated in detail. Then, integrated algorithms, namely Gradient Boosting Decision Tree (GBDT), eXtreme Gradient Boosting (XGBoost) and Random Forest (RF), are used to improve the prediction accuracy of a single machine learning algorithm. Finally, Grid Search (GS) optimization algorithm is used to further improve the generalization performance of the model. The results show that the R2 in training set of the single models is about 0.8, which is obviously lower than that of ensemble models (>0.95). It is encouraging that after conducting grid search optimization, the prediction accuracy of ML models can be improved obviously, of which R2 increasing by around 15%. In addition, the GS-XGBoost model achieved the most excellent prediction accuracy and generalization performance, with R2 values of over 99% in both the training and test sets. Finally, based on the results obtained, a software platform with a convenient graphical user interface (GUI) has been developed, which can greatly improve the efficiency of material design and testing required by researchers and engineers.