Prediction Analysis of Pre-Camber for Continuous Girder Bridge Cantilever Casting Construction Based on DBO-CNN-BiLSTM-Attention Neural Network

During the cantilever casting construction process of continuous girder bridges, it is crucial to accurately predict the pre-camber of each cantilever segment to ensure smooth closure of the bridge, structural safety, and construction quality. However, traditional methods for predicting pre-camber h...

Full description

Saved in:
Bibliographic Details
Main Authors: Jinyang Zhang, Haiqing Liu, Xiangen Gong, Ming Lei, Zimu Chen
Format: Article
Language:English
Published: MDPI AG 2025-06-01
Series:Buildings
Subjects:
Online Access:https://www.mdpi.com/2075-5309/15/13/2159
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:During the cantilever casting construction process of continuous girder bridges, it is crucial to accurately predict the pre-camber of each cantilever segment to ensure smooth closure of the bridge, structural safety, and construction quality. However, traditional methods for predicting pre-camber have limited accuracy and primarily handle linear relationships. Therefore, this paper proposes a pre-camber prediction model based on a Convolutional-Bidirectional Long Short-Term Memory network with a fusion attention mechanism (CNN-BiLSTM-Attention) and utilizes the Dung Beetle Optimizer (DBO) algorithm to optimize the hyperparameters of the CNN-BiLSTM-Attention model to enhance its predictive performance. The research results indicate that compared to several other prediction models, the model proposed in this paper demonstrates superior performance in predicting the pre-camber of continuous girder bridges. Compared to other prediction models, the evaluation metrics MAE, RMSE, and MAPE of the model proposed in this paper are minimized to 2.76 mm, 3.47 mm, and 0.70%, respectively. Applying the model proposed in this paper to the cantilever casting stage of the elevated continuous girder bridges in Shenyang Metro, China, enables pre-camber prediction with an accuracy of an average absolute error of less than 2 mm, providing a new efficient method for pre-camber prediction in cantilever casting construction.
ISSN:2075-5309