The utilization of Recycled Coarse Aggregate (RCA) in concrete has gained signifi-cant traction due to its environmental and economic advantages. However, ensuring the quality of RCA poses challenges as it is influenced by various unpredictable factors includ-ing the high water absorption of RCA, ineffective recycling processes, and the presence of contaminants. The existing body of research on the influence of RCA on the Compressive Strength (CS) of concrete has yielded inconsistent findings, and limited knowledge exists regarding the specific combination of parameters that enable effective control over CS. To address this gap, the present study aims to identify the essential parameters that contribute to controlling CS in concrete through the development of a predictive model. By investigat-ing these crucial parameters, this research intends to extent current knowledge on optimiz-ing the use of RCA in concrete.
To investigate the impact of the crucial parameters on the CS of concrete when uti-lizing RCA, a series of experiments were conducted. The RCA was obtained through the selective demolition recycling technique. The content of RCA was divided through manual separation into unbound stones, Low-Quality Recycled Aggregate (LQRA), and contami-nants. LQRA is composed of residual mortar and stones with mortar attached to their sur-face. The experiments included physical properties tests and optimization of the concrete mix designs. Additionally, relevant literature was consulted to identify the parameters that would serve as variables in constructing the predictive model. Through analysis via re-sponse surface methodology, a predictive model was developed to assess the impact of these critical parameters on the CS of concrete.
The experimental findings confirmed the statistical significance of the predictive model in assessing the impact of critical parameters on the CS of concrete. The level of LQRA was found to have a negative impact on the quality of RCA. The water-to-cement ratio was identified as a significant factor affecting the CS of concrete, with lower ratios yielding higher CS. When using RCA with high LQRA content (up to 65% of the total weight of RCA) as a substitute for natural coarse aggregate, higher replacement ratios re-sulted in lower CS.
In order to further validate the predictive model, Artificial Neural Network (ANN) modelling was incorporated as a non-linear method of assessing the relationship between the variables and the output, which is the CS. The high R2 values obtained from the ANN model demonstrated the robust alignment between the model and the data, strengthening its reliability. The integration of a Pareto chart and model-fitting regression gives a better physical understanding of the results of the predictive model by identifying influential terms and reducing complexity. The resulting model improves interpretability and predic-tive accuracy. The analyses emphasize the significance of integrating ANN and the Pareto chart approach in enhancing model validation and simplification.
These findings offer valuable insights into the parameters that are crucial to the CS of concrete which consists of RCA. By implementing the procedures that assess the quality of RCA, sustainable construction practices can be promoted, and the wider application of RCA can be facilitated on an industrial scale.