The application of a damage detection method using Artificial Neural Network and train-induced vibrations on a simplified railway bridge model

• Bridge damage location and extent are detectable using ANN with vibration responses.
• General reliability of damage detection mainly depends on the step of location.
• Damage near the middle span is easier to be detected than that near the support.
• The reliability of detection is sensitive to noise, train speed and axle load.
• Impact of noise is alleviated when resonance occurs with certain train properties.

A damage detection algorithm based on Artificial Neural Network (ANN) was implemented in this study using the statistical properties of structural dynamic responses as input for the ANN. Sensitivity analysis is performed to study the feasibility of using the changes of variances and covariances of the dynamic responses of the structure as input to the ANN. A finite element (FE) model of a one-span simply supported beam railway bridge was developed in ABAQUS®, considering both single damage case and multi-damage case. A Back-Propagation Neural Network (BPNN) was built and trained to perform damage detection. A series of numerical tests on the FE model with different vehicle properties was conducted to prove the validity and efficiency of the proposed approach. The results reveal not only that the ANN, together with the statistics, can correctly estimate the location and severity of damage, but also that the identification of the damage location is more difficult than that of the damage severity. In summary, it is concluded that the use of statistical property of the structural dynamic responses as damage index along with the Artificial Neural Network as tool for damage detection for an idealized model of a bridge is reliable and effective.