The aim of this project is the development, testing and deployment of AI algorithms for an autonomous system to operate multimode TBM’s on a tunnelling megaproject. The Autonomous TBM system is a “plug and play” solution imagined as a custom software installed on an industrial PC module. Teams developped this system as an extent of a previous technological innovation, the Variable Density (VD) TBMs which are capable of multimode operations, including full slurry and Earth Pressure Balance processes. The sophisticated operation of the multi-mode VD TBMs necessitates re-training even for experienced operators as it involves new operating techniques, providing the initial motivation to develop the « Autonomous TBM » system.
Global tunnelling boom and the bourgeoning of tunnelling projects worldwide lead to an ever-increasing demand for TBM operators. The safe delivery of tunnelling projects relies heavily on TBM operators’ competency. However, it is difficult to gauge and validate an operator’s skills through formal hiring processes. There is also a huge variability in operator competency levels and additional training requires significant resources and time. Indeed, a TBM has over 400 sensors sampling thousands of data points every minute. A computer is highly suited to process these data points and respond instantaneously with appropriate decisions. In contrast, human operators are limited by their ability to only view each data point individually and sequentially, resulting in slower response times.
The « Autonomous TBM » system has been developed in Malaysia, for the Klang Valley MRT Line 2 project, currently the largest mega-infrastructure project being undertaken by the Malaysian Government. It has already successfully completed 3km of urban tunnelling in complex geologies undercrossing sensitive infrastructure including live rail lines and a 14-lane highway.