Abstract:
This is an article in the field of ceramics and composites. To achieve the resource utilization of fly ash and accurately assess the compressive strength of fly ash concrete, three predictive models for compressive strength were constructed using machine learning modeling techniques, including traditional linear regression, decision tree, and support vector machine models. These models were utilized to model and analyze the compressive performance of the concrete. Firstly, a corresponding experimental database was established, with seven input parameters including cement, fly ash, water reducer, coarse aggregate, fine aggregate, water, and curing age, and the compressive strength as the output parameter. Based on 10-fold cross-validation, the performance of the three models on the training set was evaluated using root mean square error (RMSE), mean absolute error, and correlation coefficient, and their performance on the test set was compared. The results showed that curing age had a high correlation with compressive strength (0.60), and the correlation of fly ash with compressive strength was higher than that of cement. The traditional linear regression model exhibited an RMSE of 7.27 and 5.91 on the training and test sets, respectively. The decision tree model showcased an RMSE of 2.72 and 9.23 on the respective sets, while the support vector machine model yielded an RMSE of 5.34 and 4.09. Overall, the support vector machine model exhibited good accuracy and robust performance in predicting the compressive strength of fly ash concrete. This research can provide strength design guidance for concrete using fly ash and promote the resource utilization of fly ash.