부산대학교 도서관

The rice husk ash (RHA) concrete minimizes greenhouse gas emissions while also solving the issue of agricultural waste disposal. Though, the compressive strength prediction of rice husk ash concrete has become a new task. This research proposes novel artificial intelligence models such as CatBoost, GBM, CNN, and GRU algorithms to predict the compressive strength of RHA concrete. A comprehensive concrete dataset of 1212 points with 7 input variables (cement, water, rice husk ash, fine aggregate, coarse aggregate, age, and superplasticizer) was used to train ML and DL models and compare their forecasting performance with four models. Six different statistical indicators were applied to assess the proposed models' predictive performance. This research evaluated that both ML and DL models for the compressive strength prediction of RHA concrete yielded reliable outcomes, in which R2values were overhand 0.975 and 0.925, respectively, at the training and testing phases. The GRU model demonstrated the highest level of accuracy for compressive strength among all models, achieving approximately R2=0.99for training and almost R2=0.97phases. The error values found for RMSE, MAE, and MAPE were the lowest; however, the greatest R2, d, and CE values were obvious evidence of the model's excellent performance. The research can help the development of more sustainable and environmentally friendly building materials.