基于NRS-MSVR模型的混凝土坝变形多维精准预测

以强化多输出支持向量机(MSVR)模型在混凝土坝变形多维预测中的鲁棒性和泛化能力为目标,围绕变形多维联合精准预测展开研究。在邻域粗糙集(NRS)理论的框架内发展了一种变形影响因子贡献度度量方法,实现了多影响因子约简和关键因子的提取。视关键因子为MSVR模型的输入变量,编制迭代加权最小二乘算法进行模型求解,进而建立混凝土坝变形多维预测的NRS-MSVR模型。通过工程实例,对NSR-MSVR和MSVR模型预测效果进行对比分析。结果表明,在两模型拥有高拟合精度的前提下,NRS-MSVR模型的预测精度更高,具备更强的鲁棒性和泛化能力,是混凝土坝变形多维联合精准预测的一种新方法。

Multi dimensional Accurate Prediction of Concrete Dam Deformation Based on NRS MSVR Model

With the objective of strengthening the robustness and generalization ability of multi output Support Vector (MSVR) model in multi dimensional deformation prediction of concrete dam, studies are carried out on multi dimensional combined accurate deformation prediction. A measure method of contribution degree of deformation influencing factors is developed within the framework of Neighborhood Rough Set (NRS) theory, which realizes reduction of multiple influencing factors and extraction of the key factors. The key factors are regarded as the input variables of the multi output Support Vector Regression (MSVR) model, which is solved by the iterative weighted least square method. The multi dimensional deformation prediction model of concrete dam (NRS MSVR) is established. The comparison and analysis are implemented between NRS MSVR model and the MSVR model through an engineering example. The results show that the prediction accuracy of NRS MSVR model is higher and the abilities of robust and generalization of NRS MSVR model are stronger in the premise of the fitting precision is high for two models. So, a new approach is provided for the multi dimensional accurate prediction of measuring point deformation of concrete dam.