基于GA-BP神经网络水下隧道涌水量预测及应用

根据有关水下隧道渗流涌水影响因素的研究成果及预测涌水量时选择影响因素的准则，确定了用于预测水下隧道涌水量的6个影响因子。分析了遗传算法与BP神经网络结合的可行性，并利用遗传算法优化BP神经网络的权值和阈值，从而建立了多影响因子的 GA-BP神经网络预测模型，其收敛性能好、简单可行。通过比较GA-BP神经网络和经典BP神经网络模型的预测结果，验证了前者改良了后者的局限性并提高了预测精度。

Forecast and Application on Water Inflow of Underwater Tunnel Based on GA-BP Neural Network

According to the study results in respect of factors affecting seepage and inflow of underwater tunnel and the principals for selecting the factors in forecast of water inflow，6 factors are defined for the purpose of the inflow forecast.The paper analyzed the feasibility of combination with genetic algorithm and BP neural network and in virtue of the optimized weights and thresholds of BP neural network by means of genetic algorithm，the GA-BP neural network based forecasting model with multiple impact factors is established with good convergence and availability.Depending on the comparison between the forecasting results obtained by the models of GA BP and classical BP neural networks，it is proved that the former has improved the latter’s limitation and increased the forecasting accuracy.