改进的BP神经网络在土石坝变形分析中的应用

Application of the Improved BP Neural Network Model to Earthstone Deformation Analysis

在前人工作的基础上,针对传统神经网络存在收敛速度慢、易限于局部极小等问题,提出了几点改进措施,并将其应用到某水库土石坝变形观测资料的分析中,分析成果表明,改进的神经网络可以克服传统神经网络的缺点,很好地剔除粗差,使预测的精度和学习效率都有明显提高;充分证明了这一理论具有很强的实用性。

Based on the foundation of the predecessor,this article aimed at the low convergence rate and getting into the scope around of location drawbacks of traditional neural networks,proposed several corrective measures,and applied it in the analysis of some reservoir earth-stone dam distortion observed data.Finally,the analysis achievement indicated that the improved BP neural network model eliminate the shortcomings of traditional neural networks which can reject the thick difference successfully and made the effect prediction and learning efficiency is much better.The results validated that the method had large useful value.