基于误差分级迭代法的基坑变形预测

BP(back propagation)神经网络算法在变形预测方面存在收敛速度慢、学习效率低、容易陷入局部最小值等问题,直接影响预测结果的精准性,利用误差分级迭代法优化的神经网络能够更好地降低误差,提升预测性能.通过对比分析误差分级迭代法与BP神经网络的优势,建立误差分级迭代法模型并编制误差分级迭代法变形预测程序.采用基坑工程实测数据,经过误差分级迭代法优化后神经网络的最大误差为0.96％,与径向基神经网络预测精度相比提高3.5％,利用误差分级迭代法预测基坑变形结果其精准性较高,具有一定的实用价值.

Research on Deformation Prediction of Foundation Pit Based on Error Grading Iterative Method

Aiming at the problems of BP neural network algorithm in deformation prediction, such as slow convergence speed, low learning efficiency, and easy to fall into local minimums, which directly affect the accuracy of the prediction results, the neural network optimized by the error classification iterative method can better reduce Error, improve prediction performance. By comparing and analyzing the advantages of the error classification iteration method and BP neural network, the error classification iteration method model is established and the error classification iteration method deformation prediction program is compiled. Using the actual measurement data of foundation pit engineering, the maximum error of the neural network after optimization by the error classification iteration method is 1.15%, which is 2.61% higher than the prediction accuracy of the radial basis function neural network. The error classification iteration method is used to predict the accuracy of the foundation pit deformation results Higher, with certain practical value.