Title: Machine Learning Approach for the Flexural Strength of 3D-Printed Fiber-Reinforced Concrete (3DP-FRC) Based on the Meta-heuristic Algorithm with Sensitivity Analysis
Presented By: Nima Khodadadi, University of Miami
Description: The escalating demand for concrete in construction has underscored key challenges: rising pollution, augmented energy use, and increasing complexity in concrete structures. Recently introduced, three-dimensional printing (3DP) technology offers a potential solution, enabling the construction industry to navigate these challenges more efficiently. By adopting 3DP for concrete buildings, there is a potential to prevent the necessity for formwork and enable the construction of complex structural geometries. This innovation promises reductions in construction waste, labor costs, and overall project timelines. Simultaneously, when it comes to predicting the strength of 3D-printed fiber-reinforced concrete (3DP-FRC), the Artificial Neural Network (ANN) has demonstrated its efficacy. However, the optimal configuration of ANN remains driven by intuition rather than systematic precision. Therefore, this research endeavors to integrate a meta-heuristic algorithm with a distinct ANN design. The proposed approach in this paper optimizes parameters across a feed-forward backpropagation network, employing the Mountain Gazelle Optimization algorithm in tandem with ANN. The experimental results are subsequently evaluated using R-squared, MSE, RMSE, and MAPE metrics. The refined MGO-ANN approach boasts an impressive R-squared value of 0.97. By providing more accurate data on flexural strength, this predictive model aids in reducing the need for extensive laboratory testing, thereby saving both time and cost. Moreover, the ANN-based model's Graphical User Interface (GUI) was created as a practical tool for estimating the flexural strength of 3DP-FRC for engineering problems.
Негізгі бет Machine Learning Approach for the Flexural Strength of 3D-Printed Fiber-Reinforced Concrete
Пікірлер