Strength prediction of fly ash-based sustainable concrete using machine learning techniques: an application of advanced decision-making approaches

Concrete strength prediction, especially in fly ash-enhanced mixtures, is crucial for optimizing construction materials to ensure structural integrity and sustainability. The conventional techniques are not able to capture the high complexity and nonlinearity of the relationship between influencing factors, which lead to suboptimal performance. To overcome the aforementioned challenges, the current study proposes a Hybrid Deep Learning Model (HDLM) integrating advanced methodologies like PCA, RF, GBM, CNN, and LSTM networks. The complexity of the dataset has been reduced by using PCA to preserve more than 90% of the critical information and, hence, make the model more interpretable. RF and GBM combined to improve predictive precision while CNNs and LSTMs captured spatial and temporal dependencies inherent in microstructural images and curing process data. An integrated model showed a 10–15% improvement in the value of RMSE with more than 95% accuracy in extracting features. The HDLM outperformed the traditional models as RMSE was below 2 MPa, and the precision was 20–25% better in predictions. This research provides a sound and interpretable framework to predict the strength of concrete and, in turn, offers a powerful tool for optimizing mix designs towards enhanced structural performance for construction projects.