Multi-Lane Traffic Load Clustering Model for Long-Span Bridge Based on Parameter Correlation

Abstract

Traffic loads are the primary external loads on bridges during their service life. However, an accurate analysis of the long-term effect of the operating traffic load is difficult because of the diversity of traffic flow in terms of vehicle type and intensity. This study established a traffic load simulation method for long-span bridges based on high authenticity traffic monitoring data, and an improved k-means clustering algorithm and Correlated variables Sampling based on Sobol sequence and Copula function (CSSC) sampling method. The monitoring traffic data collected through a weigh-in-motion (WIM) system was processed to generate a multi-lane stochastic traffic flow. The dynamic response of a prototype suspension bridge under a traffic load was analyzed. The results show that the traffic load can be divided into clusters with identical distribution characteristics using a clustering algorithm. Combined with CSSC sampling, the generated traffic flow can effectively represent daily traffic and vehicle characteristics, which improves the accuracy of the assessment of the loads long-term effect. The dynamic response of the bridge to different traffic flows varied significantly. The maximum and minimum vertical displacement of the main girder was 0.404 m and 0.27 m, respectively. The maximum and minimum bending stresses of the short suspender were 50.676 MPa and 28.206 MPa, respectively. The maximum equivalent bending stress and axial stress were 16.068 MPa and 10.542 MPa, respectively, whereas the minimum values were 9.429 MPa and 8.679 MPa, respectively. These differences directly influence the short and long-term evaluation of bridge components. For an accurate evaluation of the bridge operation performance, the traffic flow density must be considered.