基于SWT和SVR的重力坝变形预测研究

Deformation prediction of gravity dams based on SWT and SVR

变形是重力坝安全状况的最直接反映,应用合理的变形预测模型可以对重力坝工作性态进行准确的监控及预警。利用Hampel滤波剔除原始变形信号粗差,通过平稳小波变换(SWT)获取变形信号多尺度分量,对比分析了不同阈值函数及阈值确定方法的去噪效果。针对预处理后数据及原始数据,分别采用逐步回归分析、BP神经网络和支持向量回归(SVR)建立变形预测模型,对比分析了各模型的预测效果。结果表明:SVR模型预测效果最好;经过预处理后的数据建模预测效果优于原始数据。

Deformation is the most direct reflection of gravity dams safety,and the operating behavior of a gravity dam can be accurately monitored and forewarned by reasonable deformation prediction models.The gross error of the original deformation signal is eliminated by the Hampel filter.The multi-scale component of the deformation signal was obtained by stationary wavelet decomposition,and the denoising effect of different threshold functions and threshold determination methods were compared and analyzed.For the preprocessed and original data,the deformation prediction models were established by stepwise regression analysis,BP neural network,and support vector regression respectively,and the prediction effect of each model was compared and analyzed.The results showed that the SVR model had the best prediction results,and the prediction accuracy of the preprocessed data was better than that of the original data.