Inverse Analysis on Thermal Parameters of Lock Head Floor Based on BP Neural Network

摘要

Subjected to construction situations, engineers of some lock head projects do not perform adiabatic temperature rising test to determine thermal parameters of concrete. To solve this problem, this paper proposed an inverse analysis method of lock head thermal parameters based on BP neural networks. Firstly, combinations of concrete thermal parameters were constructed based on uniform design theory and FEM(finite element method) analyses of temperature field using these parameters were performed to generate a series of samples. Then a BP neural network was trained by these samples. After entering the measured temperature, the neural network would output the result of inverse analysis. The future temperature curve was obtained using the inverse result and was compared to the measured temperature. The result showed the reasonability and efficiency of this method, and the rate of convergence of the BP neural network was improved by uniform design, and the accuracy of this method can meet engineering requirements.