论文部分内容阅读
现行大坝安全监控技术不能按实测信号中不同频段信号特征分别选取不同监测模型进行处理,影响了大坝变形预测精度。为此,在利用小波包分解获取实测信号中的系统信号和随机信号的基础上,提出了一种基于逐步回归和GDCS-SVM的大坝变形预测组合模型,并进行了验证。工程实例表明,GDCS-SVM预测效果优于CS-SVM,而所建组合模型预测精度高于单一监测模型,具有较强的泛化能力和较好的全局预测精度,可用于大坝变形预测。
The current dam safety monitoring technology can not select different monitoring models separately according to the signal characteristics of different frequency bands in the measured signals, which affects the accuracy of dam deformation prediction. Therefore, based on the stepwise regression and GDCS-SVM, a combined model of dam deformation prediction based on wavelet packet decomposition to obtain the system signals and random signals in the measured signals is proposed and verified. The engineering example shows that GDCS-SVM is superior to CS-SVM in forecasting performance, while the combined model has higher prediction accuracy than single monitoring model, with strong generalization ability and better global prediction accuracy, which can be used for dam deformation prediction.