Analytical Assessment of the Effects of Multicategory Errors on the Completed State of a Suspension Bridge: Stepwise Multiple Regression Analysis and FEM Validation

Starting from bridge design to the construction stage, accumulation of multiple errors related to manufacturing, installation, material properties, and other factors results in deviations between the actual completed bridge state (CBS) and the designed CBS. This study aims to reveal the patterns of effects of multiple categories of errors and the setting of random error values on the actual CBS, which is critical for ensuring the quality of completion. An analytical model for solving the CBS under the influence of multiple categories of random errors is developed to explain how error parameters influence the cable shape in the CBS from mechanical and physical perspectives. Significant parameters in each error subcategory are identified through stepwise multiple regression via specific target responses. Finally, a case study is performed for a large-span suspension bridge. A total of 337 random error parameters belonging to seven subcategories are considered, namely, horizontal position errors of cable clamps (HPEC), horizontal distance errors of lower hanger points in the main beam, horizontal distance errors of nodes in the beam segments, dead weight errors of main beam segments (DES), unstrained length errors of hangers (ULEH), flexibility coefficient errors of the main towers, and elevation errors of the main towers. Variations in the CBS under the influence of 337 random error parameters are analyzed. The accuracy of the analytical solutions is validated using an FEM. The calculations in this case indicate that the results for the total error model satisfy linear relationships with those for each error subcategory model. Compared with the remaining four error subcategories, the HPEC, ULEH, and DES make greater contributions to variations in the target responses. For the first two subcategories, the number of error parameters can be reduced by 4.5%–7.6% of the total by the following criterion: the cumulative percentage of the effect of the top-ranking regression coefficients should exceed 95%. By contrast, the DES has a more uniform effect on target responses, making the reduction of the number of error parameters more problematic.