关键词: 基坑变形预测, 小波去噪, 支持向量机, BP神经网络, 趋势项预测, 误差项预测

Abstract:

The accuracy of deformation prediction of foundation pit is low by using traditional methods. As a result, the tendency item and error item sequence of foundation pit deformation are separated by wavelet; the tendency item sequence is predicted by some optimized support vector machines; the error item sequence is predicted by chaotic BP neural network. The deformation prediction results of foundation pit can be obtained by superposition of the two prediction results; and the tracing prediction can be realized by adding later monitoring data uploading. According to case study results, the denoising effect of the wavelet functions are relatively superior and the mean relative error of the prediction results are less than 2%, which verify the validity, prediction accuracy and high adaptability of the optimized support vector machinechaotic BP neural network model.

Key words: foundation pit deformation prediction, wavelet denoising, support vector machine, BP neural network, tendency item prediction, error item prediction