|Table of Contents|

Prediction of construction alignment of long span continuousrigid frame bridge with corrugated steel webs based on MEC-BP surrogate model(PDF)

长安大学学报（自然科学版）[ISSN:1006-6977/CN:61-1281/TN]

Issue:
2021年6期
Page:
53-62
Research Field:
桥梁与隧道工程
Publishing date:

Info

Title:
Prediction of construction alignment of long span continuousrigid frame bridge with corrugated steel webs based on MEC-BP surrogate model
Author(s):
LU Zheyuan1;  WANG Xiaoming2;  ZHAO Baojun3;  REN Wanpeng4
(1. School of Highway, Changan University, Xian 710064, Shaanxi, China; 2. Engineering Research Center for〖JP〗Large Highway Structure Safety of the Ministry of Education, Changan University, Xian 710064,Shaanxi, China; 3. Shaanxi Communications Holding Group Co., Ltd, Xian 710009, Shaanxi, China;4. Shaanxi Road & Bridge Group Co., Ltd, Xian 710075, Shaanxi, China)
Keywords:
bridge engineering;  continuous rigid frame bridge;  agent model;  corrugated steel web;  configuration prediction
PACS:
-
DOI:
-
Abstract:
In order to reduce the influence of the shear deformation of corrugated steel webs on the cantilever construction process of corrugated steel webs bridge and improve the computational efficiency of the solid finite element model in the analysis of this type of bridge, a BP neural network (MECBP) proxy model based on the thought evolutionary algorithm (MEC) was proposed. An agentassisted configuration prediction method for the cantilever construction of CSWPC continuous rigid frame bridge was established to achieve high precision approximation and fast feedback for the construction state of the bridge. Firstly, the parameter sensitivity analysis method was used to identify and analyze the configuration control parameters of bridge cantilever construction, and the key response parameters were obtained to improve the efficiency of model analysis. Then, the sensitive parameters affecting the line shape were taken as input variables. Based on the parameter uncertainty, the fine solid finite element model and field test data were combined to obtain the training samples. Finally, the MEC algorithm was used to perform the “similartaxis” and “dissimilation” operation on the training sample group to get the winning subgroup. The initial weights and thresholds of BP neural network were optimized, and the training and testing of BP neural network were carried out. The MECBP configuration prediction proxy model was obtained through continuous iteration until the error requirements were met. The method was applied to the configuration prediction of the first longspan CSWPC continuous rigid frame bridge in Shaanxi Province, and the model prediction results were compared with the field measured values. The results show that model predicted values are in good agreement with the actually measured values, the predicted results can satisfy the requirements of error, and can reflect the configuration error change law of MECBP model than traditional BP model, in the bridge configuration prediction has better generalization ability, and can decrease the cost of agent model of training, the method in CSWPC continuous rigid frame bridge construction control achieve preferable application effect. 3 tabs, 12 figs, 28 refs.

References:

Memo

Memo:
-
Last Update: 2021-12-14