The article presents a study of the application of artificial intelligence algorithms in predicting the risk of early cracking in massive reinforced concrete structures using monolithic foundation slabs as an example. The current experience of using algorithms such as convolutional neural networks, deep learning tools (YOLOv5 model) for crack detection at various stages of the life cycle of massive reinforced concrete structures is analyzed. The causes of crack formation, physical and mechanical processes, including cement hydration are considered.
A model has been developed that predicts the magnitude of the tensile stress level in monolithic foundation slabs during construction, based on CatBoost using Python, allowing to predict the risks of early cracking with an accuracy of up to 98%.
The model was trained on synthetic data containing various design parameters and material properties, including the geometric dimensions of the slabs, the temperature on the upper surface, the heat transfer coefficient on the upper surface, the curing rate, the class of concrete and the characteristics of the soil base. Statistical analysis of the data was performed, a correlation matrix was constructed. Practical and predicted values of the model were visualized in the form of a scatter plot. The most significant parameters influencing the risk of early cracking in massive monolithic foundation slabs were obtained. The constructed model passed quality assessment according to three metrics: MAE=0.0011; MSE=4.038; MAPE=0.0014.