Damage identification of bridges from signals measured with a moving vehicle

Abstract

Identifying damage of a bridge from a vehicle moving over it is an attractive idea especially for those bridges without structural health monitoring systems as it is faster than putting sensors on the bridges. Many parts of highways and railways have been constructed on bridges and it is important to ensure that they are in good conditions. Therefore a large amount of bridges need to be monitored and for the sake of economy the monitoring should be efficient. If an instrumented vehicle can identify the occurrence and locations of damage by running over the bridges, it would save a lot of labor and time. As acceleration is easier to acquire, it is used as the main signal for damage detection. Research in this area is relatively little, not to mention the need to take into account road surface roughness and experimental verification.

Frequencies can be conveniently extracted from the vehicle response. The damage can hence be identified based on the relationship between the change of frequencies and the fractional change of strain energy. A vehicle-bridge interaction system is used to simulate the process of a vehicle running over a bridge and obtain the vehicle response for investigation. The proposed method can identify damage of simply supported and multi-span continuous bridges taking into account road surface roughness and measurement noise. They are also validated in the laboratory where a simply supported bridge is modeled using an aluminum beam and the vehicle is modeled with aluminum vehicles. This method can limit the damage location to two potential locations.

The multi-level multi-pass strategy makes use of the identification from the above method, applies genetic algorithm and lets the vehicle run over the bridge at various speeds. The unique damage location can then be identified. A numerical study for simply supported bridges and multi-span continuous bridges has verified its effectiveness.

Continuous wavelet transform (CWT) can identify local changes in a signal as damage is assumed to cause local change to the vehicle response, which makes it suitable for damage detection from vehicle response. However, the road surface roughness and measurement noise often mask the information about damage. Smoothing technique and damage indicators are proposed to help with the identification. By validating the method with a numerical vehicle-bridge interaction system and model tests in the laboratory, the damage can be correctly identified. Additional masses and sinusoidal excitation force can help with the identification too.

Repeated application of CWT involves applying the CWT to the coefficients of continuous wavelet again and again, which can also improve the results. If CWT is treated as a mathematical microscope, repeated application of CWT is like amplifying the signal several times. The effectiveness of the method has been verified numerically and experimentally.

In summary, a convenient and efficient technique to test the conditions of bridges by putting sensors on a moving vehicle is proposed and the method is verified by numerical and experimental studies. It can provide an alternative or a useful complement to conventional structural health monitoring systems.