Automated machine learning techniques for estimating of elastic modulus of recycled aggregate concrete: Fid-Bau Portal

Automated machine learning techniques for estimating of elastic modulus of recycled aggregate concrete

Chien‐Ta, Chen / Shing‐Wen, Tsai / Liang‐Hao, Hsiao

The utilization of recycled aggregates (RA) in producing novel concrete can contribute to the resilience of the building sector. However, it is important to thoroughly evaluate the mechanical properties of this variety of aggregate before incorporating it into various applications. This study used Gaussian process regression (GPR) and Decision Tree (RT) to estimate the $E_{RAC}$ because the current equations for the modulus of elasticity of concrete may not apply to recycled aggregate concrete (RAC) concrete. On the other hand, the Dwarf mongoose optimizer (DMO) and Phasor particle swarm optimizer (PPSO) were combined with related models. They formed hybrid models to improve the accuracy of developed models. In this study, the hybrid models were evaluated and compared in three phases, which 70% of the samples for training, 15% for validation, and the remaining 15% for testing phase. In addition, several statistical evaluation metrics were employed to assess the precision and effectiveness of the established models. The performance of the models was compared with error metrics and coefficient correlation to obtain a suitable model. The results generally indicate that the PPSO algorithm showed a more acceptable performance than other algorithms coupled with models. In general, GPR‐PPSO can obtain $R^{2} = 0.995$ and $RMSE = 0.423$ with 0.62% and 32% difference than RT‐PPSO.