Damage Detection of Steel Structures Based on Gated Recurrent Unit Neural Network Using Acceleration Data

Abstract

Health monitoring of a structure is very important for early detection of any changes that occur in that structure which may hamper its overall safety and serviceability. Due to several limitations of traditional health monitoring, damage assessment of any structure using Artificial Intelligence (AI), precisely by Deep Learning (DL), is gaining importance with the advancement of technology. A detailed comparison of different DL algorithms is performed in this study for damage detection using time-domain acceleration data, where Gated Recurrent Unit Neural Network (GRUNN) clearly outperforms other DL algorithms. Here, a novel two-stage damage detection method has been proposed for the localization and quantification of damage in steel structures using GRU neural network. Simulated acceleration data of a simply supported steel beam, generated from finite-element based software ABAQUS is used in the proposed algorithm to detect the location and severity of damage. Varying percentages of Gaussian random noise are added to the acceleration data to generate noisy simulated data that resemble the practical scenario. Finally, the algorithm is validated by detecting different damage scenarios from real-life raw acceleration data of the Old ADA bridge to check the robustness of the model. The investigation concludes that the proposed two-stage damage detection method using GRU is highly effective in localizing and quantifying damages of steel structures from the time domain acceleration dataset.