الفهرس Only 14 pages are availabe for public view
 |  | from | 349 |  |  |
| | | from | 349 | | |
Abstract
Tunnels are often designed using uncertain geotechnical data, which can impede precise prediction of the associated deformations, and lining stresses. As a result, engineers usually opt to re-appraise the assumed parameters by inverse analysis using monitoring measurements. The parameters obtained from the back-analyses can then be implemented in subsequent geotechnical assessments. In this research, Artificial Neural Networks (ANNs), one of the artificial intelligence techniques that simulate the human brain, are utilized to
simulate the inverse analysis of the tunnels.
A Staged approach was adopted in the current research. Firstly, a numerical
model using the finite element (FE) method was developed to estimate the
ground deformations associated with the varying ranges of various design
parameters through a parametric study. Subsequently, ANNs were used to
develop a computational model that was trained in the inverse direction to
evaluate the most influential parameters on the tunnel design. The training was performed using the deformations obtained by the numerical model. The proposed FE-ANN integrated model was also verified against a wellacknowledged closed-form/ graphical solution for the problem of stress relief associated with a circular tunnel in granular soils.
Furthermore, a real case study that presents a tunnel constructed by the New
Austrian Tunnelling Method (NATM) in weak rocks was used to check the
efficiency of this new integrated model. Applying the new integrated model
revealed the capability of the new model to estimate the most influential
soil/rock parameters of the tunnel design with high level of accuracy as well as the model factors representing the construction sequence. The illustrated results of the proposed approached confirm that the parameters that were inversely estimated, substantially enhanced the model and decreased the difference between the predicated and the observed deformations. In conclusion, the
present study demonstrated that the new integrated model evidently can be a
useful technique for the inverse analysis of tunnels and may surpass the
traditional back-analyses techniques by its simplicity in formulating the
analyses and its ability to accommodate observations error.