Prediction and sensitivity analysis of self compacting concrete slump flow by random forest algorithm

Abstract

Self-compacting concrete (SCC) is a construction material with many advantages, including high performance and the capacity to self-compact without mechanical vibration. As a result, SCC is widely used in construction, especially at locations where concrete structures are difficult to construct. Filling ability is one of the three basic requirements that must be met when designing the SCC mix. The slump flow (SF) is used to determine the SCC mixture's filling capacity. As a result, it is critical to estimate this number fast and precisely. The purpose of this study is to propose the use of a random forest (RF) model to predict the SF of SCC and to assess the effect of input parameters on output parameters. The study constructed the RF model using a dataset of 507 experimental results collected, which is the biggest data collection compared to previous studies on this subject. Additionally, a 10-fold cross-validation approach is used to improve the model's prediction performance. As a result, the performance assessment criteria for the testing dataset have values of RMSE = 59.5664 mm, MAE = 32.4483 mm, and R = 0.8614, respectively. This result shows that the RF model is an effective tool in predicting the SF of SCC.