Study on the Application of the Back-propagation (BP) Neural Network to Predict the Performance of Recycled Brick Aggregates

The rapid advance of urbanization leads to growing construction waste. To achieve the goals of "peak carbon emissions and carbon neutrality", construction waste must be treated properly to reduce the consumption of non-renewable resources. In this paper, five general physical parameters that affect the compaction performance of the recycled material-the proportions of fine, medium, and coarse particles, moisture content, and the brick-concrete ratio-are selected to establish a prediction model based on the BP neural network through the BP neural network algorithm, based on the array test of a metro depot project. The results demonstrate that the relative errors between the fitted and predicted values of the BP neural network and the corresponding measured values fluctuated within small ranges, which indicates that it is feasible to apply the BP neural network to predict material performance.