Bridge influence line identification based on big data and interval analysis with affine arithmetic

Abstract On-site bridge influence line (IL) calibration traditionally relies upon attached sensors and the pre-weighing of trucks, which limits its popularity for engineering applications. This paper proposes a novel technology for bridge IL identification based on big data of pre-defined statistical vehicle information and interval analysis using the affine arithmetic (AA) method. By employing computer vision techniques, the individual axle-weight intervals were determined. The AA algorithm was used to obtain the interval solutions due to its high computation efficiency (only consumes 1 s). The binary classification algorithm for the support vector machine was adopted to extract the certain IL from the massive amount of IL interval data. The proposed method was verified by employing vehicle-bridge coupled numerical simulation and laboratory tests. Calculations results indicated that the errors of the distinguished IL obtained by the proposed method ranged from 7.43% to 15.16%. The ILs were successfully identified in multiple cases.