基于IGOA-ELM的拱桥多节段吊装扣挂施工线形预测方法

A Method for Predicting Geometric Shape in Arch Bridge Multi-segment Hoisting and Buckling Construction Based on IGOA-ELM

为了实现对大跨度钢管混凝土拱桥扣索索力的高效率优化,提出了一种基于改进的蝗虫优化算法(IGOA)和极限学习机(ELM)的线形预测方法和索力优化模型。采用非线性递减函数和精英保留策略改进蝗虫优化算法,使其适应高维优化问题。建立了基于改进蝗虫算法优化极限学习机的拱桥线形预测模型,将有限元模型的计算结果作为极限学习机的训练样本并记忆最佳训练参数,建立了基于IGOA-ELM算法的索力优化模型。以某大跨度钢管混凝土拱桥为工程背景进行了应用研究,根据实测线形结果对优化模型结果进行了验证。结果表明:IGOA优化后的极限学习机相较标准ELM、BP神经网络和SVM模型预测精度更高;优化输入权值矩阵和隐含层偏差后的ELM模型可以精确预测不同索力组合下的成拱线形,对数据样本的泛化能力较强,对于样本索力组合的预测结果十分接近有限元模型的实际计算结果;IGOA算法对高维优化问题具有较好的收敛性,与标准PSO、标准GOA算法相比,IGOA算法均能收敛至不同测试函数的理论最优值附近,算法性能得到大幅提升;基于IGOA-ELM算法优化得到的扣索张拉力相较设计张拉力有效改善了成拱线形,达到了一次斜拉扣挂下的线形优化目标;与基于传统有限元法进行优化的方式相比,IGOA-ELM优化耗时最短,有效提高了结构优化效率。

In order to optimize the cable force of long-span CFST arch bridge efficiently,a geometric shape prediction method and a cable force optimization model based on IGOA and ELM are proposed.The IGOA is improved by using nonlinear decreasing function and elite retention strategy to adapt it to high-dimensional optimization problems.A model for predicting arch bridge geometric shape based on IGOA-ELM is established.Taking the calculation result of the finite element model as the training samples of the ELM and remembering the best training parameters,a cable force optimization model based on IGOA-ELM algorithm is established.The application research is conducted based on a long-span CFST arch bridge,and the optimization model result is verified according to the measured geometric shape.The result shows that(1)The prediction accuracy of the ELM optimized by IGOA is higher than that of the standard ELM,BP neural network and SVM model;(2)After optimizing the input weight matrix and hidden layer deviation,the ELM model can accurately predict the arch geometric shape under different cable force combinations,it has strong generalization ability for data samples,and the prediction result for sample cable force combinations are very close to the actual calculation result by the finite element model.(3)IGOA algorithm has good convergence for high-dimensional optimization problems.Compared with the standard PSO and standard GOA algorithms,IGOA algorithm can converge near the theoretical optimal values of different test functions,and the algorithm performance is greatly improved.(4)Compared with the design tensioning force,the cable buckling tensioning force optimized by IGOA-ELM algorithm effectively improves the arch geometric shape,and achieves the geometric shape optimization goal under single cable-stayed buckling.(5)Compared with the optimization method based on traditional finite element method,IGOA-ELM optimization takes the shortest time and effectively improves the efficiency of structural optimization.