Analysis and prediction of compressive strength of calcium aluminate cement paste based on machine learning

Calcium aluminate cement (CAC) is an important hydraulic cementitious material. It is widely used in construction, metallurgy, chemical industry and other fields due to its high early strength. The factors affecting its strength are also very complex. The research focus of this paper is to establish a prediction model for the compressive strength of CAC paste, so as to assist scientific research and practical engineering to quickly predict the strength of CAC paste at different ages under different mix ratios and curing conditions. In this paper, 273 sets of data are trained and tested based on support vector regression (SVR), random forest regression (RFR), gradient boosting (GB) and extreme gradient boosting (XGB) algorithms. It is found that the prediction accuracy of GB model can reach 89%. Meanwhile, based on the GB model, the feature importance analysis, global interpretation and dependence analysis are carried out. It is found that the main factors affecting the strength of CAC are relative humidity, silica fume content and curing temperature. To obtain high-strength CAC paste, the recommended mix ratio and curing conditions are as follows: Al₂O₃ content is 67%, CaO content is 32%, silica fume replacement rate is 10%, water–cement ratio is 0.1, relative humidity is 90%, curing temperature is 5 °C and low-temperature treatment time is greater than 60 days. Finally, a graphical user interface is established to facilitate direct prediction of CAC paste under new mix ratio and curing conditions.