Efficient compressive strength prediction of concrete incorporating industrial wastes using deep neural network

The prediction of concrete compressive strength based on mixing proportions using statistical and machine learning techniques has gained significant attention due to its relevance in the industrial context. However, most existing models have been developed with limited experimental data. In this study, a neural-based prediction model is proposed that employs both deep neural network (DNN) and artificial neural network (ANN) approaches to accurately forecast the compressive strength of high-strength concrete using eight input parameters. To ensure the robustness of the present model, a comprehensive dataset comprising over 1000 building site records has been used. For the development of the ANN model, MATLAB's ANN Tool is utilized and experimented with three different algorithms namely, Levenberg–Marquardt (LM), Bayesian regularization (BR), and scaled conjugate gradient (SCG). Additionally, the DNN model using Python coding is implemented. The prediction accuracy of the models is evaluated by analyzing the root mean square error (RMSE) and coefficient of determination (R2), while also employing Taylor diagram to assess their performance. The results demonstrated that the DNN model achieved remarkable accuracy in predicting the compressive strength of concrete incorporating industrial waste, yielding an R2 value of 0.972. Furthermore, a sensitivity analysis revealed that the cement content, amount of blast furnace slag, and age of concrete were identified as the most influential parameters affecting the compressive strength. This research contributes to the field by providing an effective prediction model for high-strength concrete compressive strength, leveraging the power of neural networks, and incorporating a comprehensive dataset. © 2023, The Author(s), under exclusive licence to Springer Nature Switzerland AG.