Tweet Share

• Send by email

  Send by email

  

  Modelling of surface subsidence induced by tunnelling using Artificial Neural Network

  There are several empirical, semi-empirical, analytical and numerical approaches available for predicting tunnel subsidence. Herein, new viable approaches are suggested for the prediction of ground surface settlements caused by tunnelling, incorporating artificial intelligence models. Current research is focused on developing a simple method in order to provide a predictive model to estimate surface subsidence induced by tunnelling using an artificial intelligence algorithm. The simulated model is comprised of two groups of parameters affecting tunnel surface subsidence which are the geometrical and excavation characteristics. So the tunnelling process in a part of Tehran subway projects, which are currently under operation, is simulated using “Artificial Neural Network (ANN)” algorithm. The study aimed at developing approaches that can be used with minimal changes for predicting surface subsidence induced by new tunnels to be excavated in similar conditions in the other locations. In this regard, after introducing “Artificial Neural Network (ANN)” algorithm and tunnel subsidence affecting parameters, can be used in the developed model, the case study would be simulated. Due to similarity of the geotechnical properties all over the study area, the case study was simulated based on the tunnel excavation and geometrical characteristics as two groups of the most important effective input parameters. In other words, the simulated ANN is trained using nine key parameters which consist of geometrical and excavation characteristics as the model input parameters and site monitoring results (precise levelling data) as the model output parameter. Based on the results of training step, the simulated model estimates subsidence values in a testing procedure in comparison with the observed ones.

  Recipient :

  Name ^* :

  E-mail address ^* :

  ^* Required fields

  Send   or  Cancel

• Print