基于正交设计-BP神经网络的混凝土坝参数反演分析

Research on mechanical parameters back analysis of concrete dam based on orthogonal design and BP neural network

摘要

在使用传统神经网络模型进行混凝土坝材料力学参数反演分析的基础上，将正交试验设计理论与BP神经网络模型相结合，对混凝土坝材料力学参数反演模型进行了优化。通过构建三维有限元模型，以拱坝实体工程及其变形监测数据进行了计算分析，反演得到的混凝土和基岩参数，计算结果与设计值相近且有一定程度提升。最后将反演成果作为有限元计算模型的输入，输出成果验证了计算结果的可靠性，从而证明了优化后的方法提高了神经网络模型计算的效率和精确度。

Abstract

Based on the back analysis of concrete dam mechanical parameters by using traditional neural network model, the BP neural network model is combined with orthogonal experiment design theory, the back analysis model of concrete dam mechanical parameters is optimized. By building a three-dimensional finite element model, the parameters of concrete and bedrock are calculated based on the arch dam and its deformation monitoring data by the back analysis. The calculation result is close to the design value and has a certain degree of improvement. Finally, the back analysis result are taken as the input of the finite element model, and the reliability of the calculation result is verified by the output result. It is proved that the optimized method improves the efficiency and accuracy of the neural network model.

关键词

混凝土坝 / 反演分析 / 正交设计 / 神经网络

Key words

concrete dam / back analysis / orthogonal design / neural network

Author summay

赵 骕(1969—), 男, 高级工程师, 学士, 主要从事工程勘察工作。E-mail: [e-mail]