Abstract

The assessment of stability of retaining walls that were constructed to prevent soil instability and collapse under the loads on nearby piles could be a sophisticated task. In urban areas, buildings or infrastructure are sometimes built relatively close to each other. Often pile foundations or groups of piles are used as the primary supporting systems and, inevitably, existing nearby retaining walls would be affected by these structures. The maximum wall deformation of these retaining walls was selected as a key factor to be determined to assess the retaining wall stability. In order to investigate the effect of loaded piles on the retaining wall, a set of parameters were selected such as the pile length, diameter and its location from the retaining wall. Considering all these parameters could lead to a large number of scenarios in order to establish the sensitivity of the system with respect to each variable. To reduce the required number of models needed to be analyzed, an Artificial Neural Network (ANN) was developed based on a representative dataset of base parameters. Similar to our brain, once the input (parameters) and output (maximum displacement and its location) baseline are given, the ANN is able to simulate and train by itself to provide a credible prediction of any corresponding scenario. Using the trained ANN model, for future designs engineers can predict a retaining wall maximum deformation and location under different geometrical scenarios, and as well to enhance or improve the serviceability of the entire pile-wall system.