Double Hybridized artificial network for the prediction of HPC concrete compressive strength

Abstract

The compressive strength of high-performance concrete encounters difficulties in prediction due to supplementary cementitious materials in its mix designs. There are non-linear relationships between the input materials and the compressive strength. Distinguishing these relationships is intensified through innovative mix designs of high-performance concrete. Artificial neural networks based model incorporated in the present study to narrow down the intensified difficulties of compressive strength prediction. Moreover, to improve the robustness and flexibility of the model and reduce its complexity, Grey Wolf and Ant Colony Optimization algorithms optimize the ANN model. Different statistical metrics are employed to appraise the assessment of models. Considering RMSE values, the values of ”GWANN-I ” and ”ACANN-I” are 1.6674 and 1.8653, respectively, delivering an acceptable performance in compressive strength prediction of HPC concrete. The OBJ values demonstrated that the ACANN-I with the value of 1.4499 outperforms best compared to other developed hybrid models and can be introduced as the best model for HPC compressive strength prediction.