基于粒子群算法优化BP神经网络的地铁暗挖段地表变形预测

Surface Deformation Prediction of Metro Undercutting Section Based on PSO-BP Neural Network Algorithm

在使用传统BP神经网络算法建模进行预测过程中,由于初始权值和阈值是随机给定的,易使网络陷入局部最优,从而导致预测精度较低。利用具有较强优化能力的粒子群算法(particle swarm optimization,PSO)优化BP神经网络在训练过程中的初始权值和阈值,建立新的预测模型,以青岛地铁3号线保河区间隧道监测数据为例进行验证分析,研究结果表明,与传统BP神经网络预测算法相比,使用PSO算法优化的BP神经网络预测算法可以得到更优的预测结果。

When we build models to predict using the traditional BP neural network algorithm , because the initial weights and thresh-olds are stochastic , it is easy to make the network into local optimum , finally resulting in a lower prediction accuracy .Use particle swarm optimization algorithm to optimize initial weights and thresholds of BP neural network in the training process , then establish new forecasting model , applying to Qingdao monitoring data of Metro Line 3 Paul River tunnel as example to analyze , and the results show that, compared with the traditional BP neural network prediction algorithm , using PSO algorithm to optimize BP neural network can get a better predictions .