Intelligent Diagnosis of Highway Bridge Technical Condition Based on Defect Information

In the bridge technical condition assessment standards, the evaluation of bridge conditions primarily relies on the defects identified through manual inspections, which are determined using the comprehensive hierarchical analysis method. However, the relationship between the defects and the technical condition of the bridges warrants further exploration. To address this situation, this paper proposes a machine learning-based intelligent diagnosis model for the technical condition of highway bridges. Firstly, collect the inspection records of highway bridges in a certain region of China, then standardize the severity of diverse defects in accordance with relevant specifications. Secondly, in order to enhance the independence between the defects, the key defect indicators were screened using Principal Component Analysis (PCA) in combination with the weights of the building blocks. Based on this, an enhanced Naive Bayesian Classification (NBC) algorithm is established for the intelligent diagnosis of technical conditions of highway bridges, juxtaposed with four other algorithms for comparison. Finally, key defect variables that affect changes in bridge grades are discussed. The results showed that the technical condition level of the superstructure had the highest correlation with cracks; the PCA-NBC algorithm achieved an accuracy of 93.50% of the predicted values, which was the highest improvement of 19.43% over other methods. The purpose of this paper is to provide inspectors with a convenient and predictive information-rich method to intelligently diagnose the technical condition of bridges based on bridge defects. The results of this research can help bridge inspectors and even non-specialists to better understand the condition of bridge defects. © 2024 The Authors. Published by Tech Science Press.