Intelligent mix design of recycled brick aggregate concrete based on swarm intelligence

The mechanical properties of recycled brick aggregate concrete (RBAC) are significantly affected by design parameters such as recycled brick aggregate (RBA) replacement ratio and water-cement ratio, which not only increases the difficulty of designing RBAC by experimental or theoretical methods, but also makes the mechanical properties (compressive strength, flexural strength, etc.) of RBAC difficult to match the design requirements. Therefore, an intelligent mix design method for RBAC is proposed in this paper, which can determine the best RBAC mix proportion design scheme according to the design parameters of RBAC (cement content, water-cement ratio and crushed brick ratio, etc.). This method is compared with the performance of 8 machine learning algorithms by creating a database containing 8 input variables. The most suitable model for predicting compressive strength of RBAC is GWO-BP. The predicted/experimental values obtained by the model on the training set and the testing set are 0.98 and 0.96, respectively, and the determination coefficients R2 are 0.97 and 0.99, respectively. SHapley Additive exPlanations (SHAP) is used to analyze the mechanism of design parameters on the compressive strength of RBAC, and the key parameters affecting the compressive strength of RBAC are cement content (Cement) and water-cement ratio (W/C). The Multi-objective optimization (MOO) model is used to target the compressive strength, carbon emission and cost of RBAC, and the scheme with the highest score is determined as the best design scheme of RBAC mix ratio by determining the Pareto optimal solution set of RBAC. The results show that the design method proposed in this paper can effectively guide the mix design of RBAC, the machine learning model optimized by swarm intelligence can effectively predict the compressive strength of RBAC concrete, while the developed MOO model can effectively determine the optimal mix design method.