Back Analysis of Mechanical Parameters of Arch Dam Using an Improved BP Neural Network Algorithm

The backpropagation (BP) neural network algorithm was improved and applied to construct a back analysis model for the study of the mechanical parameters of an arch dam. Samples were obtained from forward analysis using the three dimensional finite element method. The obtained samples were then used totrain the network, and the measured displacements were applied to invert the mechanical parameters. The exact mechanical parameters were calculated and analysed according to the back analysis of parameters. Computational results agree well for some blocks with the measured values, indicating compliance withengineering requirements. Furthermore, the BP neural network algorithm has convergence and corresponds to dam's natural behavior. The results show that the improved BP neural network algorithm and threedimensionalfinite element method are feasible techniques for inverting the thermal parameters of arch dams

Keywords:

improved BP neural network algorithm , three dimensional finite element method , displacement , back analysis , mechanical parameter