改进的BP神经网络在路基沉降预测中的应用

Application of Improved BP Neural Network in Prediction of Settlement of Roadbed

针对传统BP神经网络存在的缺点,提出基于遗传优化的变梯度反向传播的BP神经网络预测方法,采用遗传算法优化BP神经网络的初始权重,建立路基沉降预测模型。该模型可克服BP神经网络模型存在的收敛速度慢、易陷入局部极小点等缺点。结合现场实测数据,将该优化模型与指数曲线模型、双曲线模型、灰色预测模型和传统BP神经网络预测模型对比,结果表明改进的BP神经网络在路基沉降预测中精度最高,适宜于广泛推广应用。

Addressing the shortcomings of the traditional BP neural network,the author proposes a new method of prediction by means of back-propagation（BP）neural network which is based on genetic optimization and variable gradient back propagation.The coupling model of genetic algorithm and back-propagation（BP） neural network was applied to the Prediction of Roadbed Settlement,aiming at overcoming shortcomings of the BP neural network model,such as susceptibility of falling into local minimum value and being slow in convergence.Based on field measured data,the comparisons of new model with exponential curve model,hyperbolic model,grey forecasting model and traditional BP neural network prediction model,have shown that the improved BP neural network is of highest accuracy,and adaptable to wide range of applications.