Prediction of Water Resistance of Magnesium Oxychloride Cement Concrete Based upon Hybrid-BP Neural Network

To obtain the magnesium oxychloride cement concrete (MOCC) ratio with excellent water resistance quickly and accurately, a BP neural network (BPNN) model with a topology structure of 4-10-2 was designed, and the PSO (particle swarm optimization), GWO (gray wolf optimization), and WOA (whale optimization algorithm) algorithms were used to optimize the model. The input layer parameters of the model above were n(MgO/MgCl(2)), Grade I fly ash, phosphoric acid (PA), and phosphate fertilizer (PF) content, and the output layer was the MOCC’s compressive strength and softening coefficient. The model had a dataset of 144 groups, including 100 training set data, 22 verification set data, and 22 test set data. The results showed that the PSO-BPNN model had the highest predictive accuracy among the four models, with a mean R(2) of 0.99, mean absolute error(MAE) of 0.52, mean absolute percentage error(MAPE) of 0.01, and root mean square error (RMSE) of 0.73 in predicting compressive strength, and a mean R(2) of 0.99, MAE of 0.44, MAPE of 0.01, and RMSE of 0.62 in predicting the softening coefficient. The results showed that using the PSO-BPNN to predict the compressive strength and softening coefficient of MOCC is feasible and can provide theoretical guidance for designing the MOCC mix.