Compared with other means of transportation, underwater tunnel has its own advantages, so it has been developed quickly both at home and abroad in recent years. However, unlike mountain tunnel, underwater tunnel has its own characteristics, such as high water pressure, abundant water source and lacking natural drainage. All these characteristics lead to more serious leakage problems, which are difficult to be dealt with comparison to land tunnel. Therefore, According to the specificity of underwater tunnel, a reasonable prediction of water inflow becomes the key factor of the underwater tunnel waterproof design and construction.The methods to forecast the tunnel inflow directly influence the accuracy of it. This thesis uses genetic algorithm and BP neural network (GA-BP) method to predict the water inflow of Guangzhou Shiziyang tunnel. Also, this work is the technology development project supported by the Ministry of Railways.The research is mainly focused on two aspects.(1) According to domestic and foreign research achievements related to influencing factors of water inflow into underwater tunnel and engineering experiences, the influencing factors of the Shiziyang tunnel are analyzed. Together, under direction of the basic rules when predicting tunnel inflow, six factors are identified in this paper.(2) The feasibility of the combination of genetic algorithm and BP neural network is analyzed before the optimization for weights and thresholds of BP neural network, and then the prediction model for underwater tunnel inflow is established.Meanwhile, by comparing the prediction results of GA-BP and BP neural network, the facts that the former method modifies the limitations of the latter one and improves the prediction accuracy.The results show that this model is simple and feasible and has good performance in convergence. Also the method offers a new way in forecasting water inflow of underwater tunnel. |