A novel bridge damage detection method based on the equivalent influence lines – Theoretical basis and field validation

Abstract

This paper presents equivalent deflection as a new concept for bridge damage detection. The equivalent deflection is similar to the real deflection and can be inferred from bridge acceleration measurements (accelerometer). While calculating measured acceleration, the method essentially explores the static response of the bridge to load. It is laid on the premise that most of the deflection response is generally static rather than dynamic. By using the equivalent deflection from a group a trains without knowing their bogie weights, the average single bogie equivalent deflection response (ASBED) can be calculated and used as an indicator of structural damage. Initially, the theoretical procedure is described to find the equivalent deflection from an acceleration signal using a moving force identification algorithm. Then, numerical analyses are carried out to validate the algorithm. The results confirm that the equivalent deflection is close to the real deflection and contains only a small dynamic component. Then, a method to find the shape of influence line is developed. Using real acceleration data from a 26.8 m simply supported railway bridge (located over the West Coast Mainline in Staffordshire, UK), the shapes of equivalent influence lines are obtained from different groups of trains with very good repeatability. The ASBED response concept is introduced as the product of the ‘inferred bogie weight’ and the shape of the influence line. Good repeatability of the ASBED response is obtained from site with variability less than 1.5%. Theoretically, this ASBED response should be able to detect changes in the influence line when the average bogie weight at the site is repeatable. The last section explores the changes in the theoretical influence line due to different bridge damage conditions by using a 2D grillage bridge model. The results show that a 2% overall reduction in material stiffness or 20% local reduction (over 1.5 m) should be detectable using this method.