PREDICTION OF CONCRETE COMPRESSIVE STRENGTH USING ANN BY THE HELP OF REGRESSION ANALYSIS

Abstract

Regression analysis's interpretability and Artificial Neural Networks' (ANN) capacity are combined in this study to provide a hybrid technique for forecasting concrete's compressive strength. The fundamental components of an extensive dataset are the design parameters for concrete mixes and the related values of compressive strength. Careful preparation is performed to guarantee the quality of the data. In developing a hybrid model, an ANN is first constructed, and then regression analysis is used to pinpoint and highlight the most important input parameters, therefore improving the model. The hybrid model improves accuracy and robustness by striking a compromise between the complexity of ANN and the ease of regression through methodical tuning. To optimize the design, the effects of different hyperparameters on model performance are rigorously examined. The resulting model's performance is evaluated against standalone ANN models and conventional regression models, and it undergoes thorough validation on an independent dataset. The hybrid model outperforms its competitors in terms of accuracy and interpretability, according to the results, and it provides a useful tool for estimating concrete compressive strength in practical applications. This novel method closes the gap between state- of-the-art machine learning techniques and conventional statistical approaches, offering structural engineers and researchers in the field of concrete technology a workable answer.