BP小波神经网络在边坡位移预测中的应用

用优化的BP小波神经网络建立的预测模型对吊钟坝边坡的变形量进行了预测。引入了共轭梯度反向传播算法来优化BP小波神经网络,从而使网络计算量大为减少,避免了网络产生局部最优的弊端,有效提高了网络的质量。与未优化的BP小波神经网络和BP神经网络的预测结果进行比较可以看出,优化的BP小波神经网络的预测值更接近GPS实测值。

Application of BP wavelet neural-network in prediction of slope displacement

Based on the engineering fact of Diaozhongba slope,an optimization method of BP wavelet neural-network was used to pre- dict displacement of slope. The method used conjugate gradient back propagation algorithm optimizing BP-WNN parameters,reduced large amount of network calculation volume,avoided the drawbacks of local optimum result and improved the quality of the network. Compared with non-optimized BP wavelet neural-network,the predicted values measured by the optimized BP wavelet neural-network were more close to those measured by GPS.