响应面法在大跨轨道斜拉桥索塔长期变形预测中的应用

Application of response surface method in long-term deformation prediction of long-span railway cable-stayed bridge tower

为探究大跨径轨道专用斜拉桥索塔长期变形的离散性,以重庆蔡家嘉陵江大桥为工程实例,考虑5个主要结构参数,进行中心复合设计法三水平试验设计,构造不含交叉项的二次多项式响应面模型,将索塔变形与各随机参数之间的复杂隐性关系通过近似的显式函数关系表达出来,运用F检验和修正的决定系数进行响应面模型的显著性和精度检验,采用后退回归进行基函数显著性分析,实现响应面模型的简化.最后,基于Monte Carlo抽样分析,对索塔长期变形进行概率意义上的预测.预测结果表明:确定性方法和不确定性方法计算所得的位移差异由1年的13.22 mm增大到30年的71.62 mm,参数的随机性对斜拉桥索塔塔顶位移的预测结果有较大影响,考虑参数的随机性有助于衔接结构设计状态与施工状态.

In order to explore the discreteness of long-term deformation of cable tower,taking Chongqing Caijia Jialing River Bridge as project background,five main structure parameters and central composite design method were adopted to conduct experiment design with three levels. Response surface models without considering the influence of cross-terms were constructed to express the complex implicit relationship between cable tower deformation response and random parameters by approximate explicit functions. Significance test and accuracy test on response surface models were carried out using F test and corrected determination coefficient. To simplify the response surface models,the significance of basis function was analyzed by backward regression. The long-term cable tower deformation prediction in the sense of probability was conducted by Monte Carlo sampling analysis. The forecast results show that the displacement difference calculated by the deterministic method and uncertainty method is increased from13. 22 mm in one year to 71. 62 mm in 30 years. The random parameters have great influence on the prediction of tower displacement of cable-stayed bridge,and considering randomness of parameters is suitable to connect structure design and construction.