ABSTRACT
A new concept, which combines machine and monitoring data analysis with numerical simulation of mechanized tunneling processes, is presented in this work. The main idea is to fuse sensors data collected from the TBM and the settlement measurements with finite element simulations capturing the physical behavior of the tunneling process. Both data and simulation based prediction models are developed to assist the TBM steering with respect to the settlement behavior in real-time. The real-time machine and monitoring data analysis offers the possibility of detecting anomalies, e.g. in the soil layers, which are not considered a priori in the simulation models. First results show that most anomalies created by concrete walls in the ground can be detected up to 1 m ahead of the cutting wheel by using the specific torque (a combination of torque and penetration rate data) together with time series preprocessing and drift detection methods.
