4.5 Article

Predicting compressive strength of green concrete using hybrid artificial neural network with genetic algorithm

Journal

STRUCTURAL CONCRETE
Volume 24, Issue 2, Pages 1980-1996

Publisher

ERNST & SOHN
DOI: 10.1002/suco.202200034

Keywords

artificial neural network; compressive strength; feature selection; genetic algorithm; green concrete

Ask authors/readers for more resources

This paper investigates the prediction of green concrete compressive strength using a hybrid artificial neural network with genetic algorithm. By constructing new parameters, using them as input variables, and performing feature selection, the performance of the prediction model is improved. The results demonstrate that the hybrid model with genetic algorithm and artificial neural network has the best prediction performance for green concrete compressive strength.
With the growing usage of supplementary cementitious materials (silica fume, fly ash, and ground blast furnace slag, etc.) in concrete, accurate prediction of green concrete compressive strength (CS) has become an issue worth investigating. In this paper, a green concrete CS prediction model was developed using hybrid artificial neural network with genetic algorithm (GA-ANN) based on 2479 green concrete CS experiment results. Also two prediction models based on support vector regression (SVR) and ANN algorithm were developed for comparison. Five new parameters were constructed based on the original nine influencing parameters of CS, taking all 14 parameters as input variables, the influence of the constructed parameters on the model performance was studied. Feature selection was then performed based on the maximum information correlation (MIC) and sensitivity analysis results, which could improve the model by deleting several parameters. The results showed that the prediction performance of the SVR, ANN, and GA-ANN models were improved after adding new parameters, the GA-ANN model has the best prediction performance. The accuracy and robustness of the GA-ANN model were effectively improved by deleting the input variables with lower MIC and sensitivity values.

Authors

I am an author on this paper
Click your name to claim this paper and add it to your profile.

Reviews

Primary Rating

4.5
Not enough ratings

Secondary Ratings

Novelty
-
Significance
-
Scientific rigor
-
Rate this paper

Recommended

No Data Available
No Data Available