Buy article PDF
The purchased file will be sent to you
via email after the payment is completed.
US$ 35
|
Smart Structures and Systems Volume 28, Number 4, October 2021 , pages 535-551 DOI: https://doi.org/10.12989/sss.2021.28.4.535 |
|
|
|
Machine learning and RSM models for prediction of compressive strength of smart bio-concrete |
||
Hassan Amer Algaifi, Suhaimi Abu Bakar, Rayed Alyousef, Abdul Rahman Mohd. Sam, Ali S. Alqarni, M.H. Wan Ibrahim, Shahiron Shahidan, Mohammed Ibrahim and Babatunde Abiodun Salami
|
||
| Abstract | ||
| In recent years, bacteria-based self-healing concrete has been widely exploited to improve the compressive strength of concrete using different bacterial species. However, both the identification of the optimal involved reaction parameters and theoretical framework information are still limited. In the present study, both experimentally and numerical modelling using machine learning (ANN and ANFIS) and response surface methodology (RSM) were implemented to evaluate and optimse the evolution of bacterial concrete strength. Therefore, a total of 58 compressive strength tests of the concrete incorporating new bacterial species were designed using different concentrations of urea, cells concentration, calcium, nutrient and time. Based on the results, the compressive strength of the bacterial concrete improved by 16% due to the decrement of the pore percentage in the concrete skin; specifically, 5 mm from the concrete surface, compared to that of the control concrete. In the same context, both machine the learning and RSM models indicated that the optimal range of urea, calcium, nutrient and bacterial cells were (18-23 g/L), (150-350 mM), (1-3 g/L) and 2×107 cells/mL, respectively. Based on the statistical analysis, RMSE, R2, MPE, RAE and RRSE were (0.793, 0.785), (0.985, 0.986), (1.508, 1.1), (0.11, 0.09) and (0.121, 0.12) from both the ANN and ANFIS models, respectively, while; the following values (0.839, 0.972, 1.678, 0.131 and 0.165) was obtained from RSM model, respectively. As such, it can be concluded that a high correlation and minimum error were obtained, however, machine learning models provided more accurate results compared to that of the RSM model. | ||
| Key Words | ||
| concrete strength; machine learning; response surface methodology; self-healing concrete | ||
| Address | ||
| (1) Hassan Amer Algaifi, M.H. Wan Ibrahim, Shahiron Shahidan: Faculty of Civil and Environmental Engineering, Universiti Tun Hussein Onn Malaysia, 86400 Parit Raja, Johor, Malaysia; (2) Suhaimi Abu Bakar, Abdul Rahman Mohd. Sam: School of Civil Engineering, Faculty of Engineering, Universiti Teknologi Malaysia, Johor Bahru 81310, Malaysia; (3) Rayed Alyousef: Department of Civil Engineering, College of Engineering, Prince Sattam bin Abdulaziz University, Alkharj 11942, Saudi Arabia; (4) Mohammed Ibrahim, Babatunde Abiodun Salami: Center for Engineering Research, Research Institute, King Fahd University of Petroleum and Minerals, Dhahran 31261, Saudi Arabia; (5) Ali S. Alqarni: Department of Civil Engineering, College of Engineering, King Saud University, Riyadh 11421, Saudi Arabia. | ||