Estimating compressive strength of lightweight foamed concrete using neural, genetic and ensemble machine learning approaches
Article
Article Title | Estimating compressive strength of lightweight foamed concrete using neural, genetic and ensemble machine learning approaches |
---|---|
ERA Journal ID | 21082 |
Article Category | Article |
Authors | Salami, Babatunde Abiodun, Iqbal, Mudassir, Abdulraheem, Abdulazeez, Jalal, Fazal E., Alimi, Wasiu, Jamal, Arshad, Tafsirojjaman, T., Liu, Yue and Bardhan, Abidhan |
Journal Title | Cement and Concrete Composites |
Journal Citation | 133, pp. 1-16 |
Article Number | 104721 |
Number of Pages | 16 |
Year | 2022 |
Publisher | Elsevier |
Place of Publication | United Kingdom |
ISSN | 0958-9465 |
1873-393X | |
Digital Object Identifier (DOI) | https://doi.org/10.1016/j.cemconcomp.2022.104721 |
Web Address (URL) | https://www.sciencedirect.com/science/article/pii/S0958946522003146 |
Abstract | Foamed concrete is special not only in terms of its unique properties, but also in terms of its challenging compositional mixture design, which necessitates multiple experimental trials before obtaining the desired property like compressive strength. Regardless of design challenges, artificial intelligence (AI) techniques have shown to be useful in reliably estimating desired concrete properties based on optimized mixture proportions. This study proposes AI-based models to predict the compressive strength of foamed concrete. Three novel AI approaches, namely artificial neural network (ANN), gene expression programming (GEP), and gradient boosting tree (GBT) models, were employed. The models were developed using 232 experimental results, considering easily acquired variables, such as the density of concrete, water-cement ratio and sand-cement ratio as inputs to estimate the compressive strength of foamed concrete. In training the models, 80% of the experimental data was used and the rest was used to validate the models. The optimized models were selected using their respective best hyper-parameters on trial and error basis; variable number of hidden layers, number of neurons and training algorithms were used for ANN, number of chromosomes, head size, number of genes, variable function set for the GEP and GBT employed number of trees, maximal depth and learning rate. The trained models were validated using parametric and sensitivity analyses of a simulated dataset. The prediction abilities of proposed models were evaluated using the coefficient of correlation (R), mean absolute error (MAE), and root mean squared error (RMSE). For the validation data, empirical results from the performance evaluation revealed that GBT model (R = 0.977, MAE = 1.817 and RMSE = 2.69) has relative superior performance with highest correlation and least error in comparison with ANN (R = 0.975, MAE = 2.695 and RMSE = 3.40) and GEP (R = 0.96, MAE = 2.07 and RMSE = 2.80). The study concludes that the developed GBT model offered reliable accuracy in predicting the compressive strength of foamed concrete. Finally, the simple prediction equation generated from the GEP model signifies its importance and can reliably be used in estimating compressive strength of foamed concrete. It is recommended that the prediction models shall be used for the ranges of input variables employed in this study. |
Keywords | Artificial neural network; Foamed concrete; Gene expression programming; Gradient boosting tree; Lightweight concrete; Optimization |
Funder | Beijing University of Technology |
Byline Affiliations | King Fahd University of Petroleum and Minerals, Saudi Arabia |
Shanghai Jiao Tong University, China | |
University of Engineering and Technology, Pakistan | |
Imam Abdulrahman Bin Faisal University, Saudi Arabia | |
University of Adelaide | |
Centre for Future Materials | |
University of Science and Technology Beijing, China | |
National Institute of Technology Patna, India | |
Library Services |
https://research.usq.edu.au/item/wq75q/estimating-compressive-strength-of-lightweight-foamed-concrete-using-neural-genetic-and-ensemble-machine-learning-approaches
Download files
Published Version
39
total views52
total downloads2
views this month1
downloads this month