RESEARCH ON MESOSCOPIC PARAMETERS CALIBRATION OF GEOPOLYMER CONCRETE UPON BP NEURAL NETWORK

This research proposes a calibration method for mesoscopic parameters of geopolymer concrete based on the Back Propagation (BP) neural network. The nonlinear mapping relationship between macroscopic and mesoscopic parameters of geopolymer concrete was established by training the BP neural network model. The mesoscopic parameters of geopolymer concrete were predicted upon BP neural network and the predicted mesoscopic parameters were used in the particle flow code numerical simulation. The accuracy of the prediction method was verified by comparing the macroscopic mechanical parameters predicted by the numerical simulated method and by the experimental method. The research results show that the prediction of macroscopic mechanical parameters of geopolymer concrete based on BP neural network has good accuracy, and that the relative error is less than 10%. The results also show that the predicted macroscopic parameters agree well with the experimental results, which verifies the calibration effect of this method.