The availability of data recorded by devices placed in a tunnel lining and the surrounding ground can be used to get valuable information about the behavior of the lining after it is installed. In particular, in the inverse problem the measurements are employed to estimate the behavior of the lining.

In order to predict the right behavior of the tunnel the interaction tunnel-surrounding ground has to be studied. To achieve this goal numerical models are performed to take into account the construction process of the tunnel and its influence on the ground. During the excavation there is a stress redistribution inside the rock mass, which finally determines the final loads acting on the supports. It becomes of fundamental importance how to deal with the in-situ geomechanical conditions and how to include them in the model to reproduce the most important factors which govern the phenomenon.

The application of Machine Learning Techniques in combination with numerical analyses is also investigated in order to increase the performance of the method. Artificial Neural Networks are employed for the definition of a surrogate model which operates with the data of the numerical simulations and the ones monitored.

Also the use of benchmarks plays a key role in order to perfect the methodological approach developed and to validate the procedure. A simple beam has being studied and used to investigate the inverse problem, employing data from a real experiment in combination with both analytical formulations and numerical models, in order to study the relations among the different quantities. The data of the experiments are going to be used to calibrate the analytical model and to compare the results.

Contact

M.Sc. Nicola Gottardi
nicola.gottardi@rub.de
+49 234 / 32-29057