Research on construction linear prediction of highway continuous rigid frame bridge based on GA-BP algorithm

Purpose The objective of this article is to address the issue of deviations between the actual alignment and the designed alignment of long-span continuous rigid-frame bridges during construction, which arise due to complex construction site environments and variations in material parameters, ultimately affecting the closure accuracy. To tackle this challenge, the study proposes a method based on a genetic algorithm (GA)-optimized backpropagation (BP) neural network for predicting alignment variations during bridge construction, aiming to enhance construction precision and closure quality. Design/methodology/approach This study establishes a finite element model based on an actual bridge engineering project and simulates the distribution range of material parameters using a normal distribution. Characteristic values are selected from the distribution range and input into the finite element model to calculate the corresponding deflection responses. The material parameter values and their corresponding deflection responses are used as training samples to train the GA-BP neural network model. Once trained, the model can effectively predict deflection variations during the bridge construction process. Findings The GA-BP neural network method proposed in this study demonstrates superior computational efficiency and prediction accuracy compared to traditional calculation methods and finite element simulation approaches, providing more reliable guidance for construction practices. In contrast to the conventional BP algorithm, the GA-BP algorithm significantly reduces errors in bridge alignment prediction, achieving an optimal mean absolute error (MAE) of 0.16067. This method substantially enhances the accuracy of bridge alignment control, mitigates construction risks and offers technical support for improving the precision of bridge closure. Research limitations/implications The method proposed in the article is more efficient compared to finite element model calculations, effectively meeting the real-time requirements of construction. It significantly reduces the labor costs associated with construction monitoring and mitigates construction risks. Practical implications This article introduces a machine learning-based method for predicting bridge construction deflection to ensure precise alignment control. Validated with actual measurement data, the model outperforms finite element simulations in computational efficiency and speed, meeting real-time construction needs. It provides innovative insights for advancing the digitization and intelligentization of bridge construction practices. Social implications With the rapid advancement of artificial intelligence (AI), the integration of AI with various industries has become an inevitable trend. The method proposed in this article combines machine learning with civil engineering, demonstrating the efficiency and superiority of artificial neural network algorithms. This approach highlights the potential for a more effective integration of AI and civil engineering, paving the way for innovative applications in the field. Originality/value This study represents the first application of a genetic algorithm-optimized backpropagation neural network for predicting deflection in the construction of continuous rigid-frame bridges, offering a novel approach for the intelligentization of bridge construction. The proposed method not only enhances prediction efficiency but also significantly reduces prediction errors, providing a scientific basis and technical support for alignment control during bridge construction. This approach holds substantial value for practical engineering applications.