Prediction of rheological parameters of 3D printed polypropylene fiber-reinforced concrete (3DP-PPRC) by machine learning

Abstract

This paper proposes a machine learning (ML) model to predict the 3D printed polypropylene fiber-reinforced concrete (3DP-PPRC) rheological properties, in which dynamic yield stress (DYS) plays a vital role. ICAR rheometer is used to measure the yield stress of the concrete mixture, where 41 mixtures were used to compile the data. In this research, four machine-learning models have been used to predict the DYS of the 3DP-PPRC, accounting for different water binder ratios (W/B) and polypropylene (PP) fiber content. The code has been generated in Python scripts. Several ML models such as random forest (RF), support vector machine (SVM), extreme gradient boosting (XGBoost), and light gradient boosting machine (LightGBM) have been used to predict the DYS, considering 80% and 20% data for training and testing, respectively while the model’s accuracy, MSE, RMSE, MAPE, and R2 were also calculated for 3DP-PPRC. The influence of each rheological parameter in the ML-based of 3DP-PPRC, Shapley additive explanations (SHAP) are also accompanied. The outcomes proved that utilizing an ML model to estimate the yield stress of 3DP-PPRC using PP fiber is a dominant approach.