Evaluation of Displacement of an L-shaped Concrete Specimen using Recurrent Neural Networks

In engineering, most structural elements are damaged locally during fabrication or maintenance. Under different loading conditions, such localized damage will further expand into larger cracks and cause structural collapse. As a result, identifying the displacement under various loads in the structural elements is critical in the risk assessment of engineering structures. The objective of the present paper is to propose a deep-learning model to examine the displacement of L-shaped concrete specimens under loading conditions. The three state-of-the-art models such as Simple RNN, LSTM, and GRU are built and trained based on load-displacement data. The experimental results show that the R2 values obtained from the Simple RNN, LSTM, and GRU models are 0.9967, 0.6169, and 0.5291, respectively. This proves that Simple RNN is superior to LSTM and GRU in the task of predicting the load-displacement relationship.