Evaluating mechanical strength of foam concrete with recycled brick powder using advanced machine learning models

The fine aggregates were replaced with Recycled Brick Powder (RBP) and mixed with foam concrete. Its mechanical and physical properties were examined using an orthogonal design experiment (5 levels) with a 4-factor analysis. By conducting specific strength analysis, it becomes possible to determine the optimal ratio for foam concrete containing RBP. These studies aimed to explore how the mechanical and physical properties of foam concrete mixed with brick powder are affected by varying factors such as water-to-material ratio, Hydroxypropyl Methyl Cellulose ether (HPMC) concentration, addition of air bubble group, and addition of RBP. Support Vector Machine (SVM), Random Forest (RF), Back Propagation Neural Network enhanced by Particle Swarm Optimization (PSO-BP), Both backpropagation neural networks (BP) and multiple linear regression (MLR) are used. When considering forecasting accuracy and model stability, it was discovered that the PSO-BP model surpassed the performance of the other five machine learning models. The experimental results confirm that the ideal concrete mix with 0.05% HPMC and 30% RBP with 0.55 w/c ratio provides the maximum compressive strength (5.68 MPa). The PSO-BP model fitted well with the highest regression value (R2=0.961).