Evaluation of the Efficiency of Artificial Intelligence Algorithms for the Detection of Structural Damage in a Steel Arch Bridge

Structures such as bridges face various events that can affect their structural capacity and put their future performance at risk. In recent years, several bridges have experienced continuous failures due to earthquakes, overloads and poor design. Current structural assessment methods are limited and are not always able to detect damage at an early stage, which can lead to a significant increase in long-term repair costs. Therefore, this paper proposes to identify structural damage on a 59.2 meter long steel bridge using modern artificial intelligence algorithms such as Scaled Conjugate Gradient (SCG) and Levenberg-Marquardt (LM). To obtain the vibration frequencies, the Hilbert-Huang transform and the Variational Modal Decomposition (VMD) method were used. In addition, neural networks such as Levenberg-Marquardt and scaled conjugate gradient are used to identify damage. The results demonstrate correct damage detection in 14 sections of the bridge and greater efficiency using the LM algorithm. In conclusion, based on the results, a 95.8% efficiency in damage detection is observed when opting for the LM algorithm in contrast to the SCG algorithm which presents 94.1%.