Abstract
Tunnel boring machines (TBMs) have been widely used in the tunnel constructions because of the advantages including high advance rate, high quality, and environmentally friendliness. However, the low adaptability of TBM remains a serious problem. Adverse geology conditions, such as spalling, rock bursting, fault fracture zones, water inflow, and super high ground stress or rock mass hardness, would result in the delay of the construction schedule, damage of facility or even casualties of the crew, which always causes huge economic loss to engineering projects. In this paper, based on a water conveyance tunnel Yin-song project located in Jilin Province of China, a data-driven model with a comprehensive procedure to identify adverse geological conditions ahead was proposed. Firstly, a great amount of data containing the operational parameters and running state of TBM was collected. After the data mining and preprocessing work, around 12,000 TBM tunneling segments were extracted. Secondly, the values of surrounding rock classification and information of lithology obtained in geological prospecting at certain measuring points were extended to be consistency with the TBM operational data in quantity, thus the machine-geological dataset was established. According to the position of the geological hazards recorded in the project, these corresponding parts of data in the established dataset was labeled, thus these tunneling segments were turned into a group of samples to train and test the predicting model. Recurrent neural network (RNN) were selected to train this model. The experimental results showed that the proposed RNN-based adverse geological conditions predicting model performed better in this case. Hence, the proposed model could be applied to predict adverse geological conditions while tunneling, which is of great significance to the safety and efficiency of TBM tunneling.
The rapid development of mechanical manufacture, informatization management and intelligent controlling is leading tunnel constructions to an intelligent direction. TBM (Tunnel Boring Machine), a kind of highly mechanized machine, is getting more and more widely used in the tunnels built in hard rocks. Although TBM has amounts of advantages such as high advance rate, safety, environmentalfriendliness and so forth, its further application may be limited by its poor adaptability in different geological conditions. The geological structures encountered in the tunneling process are always complicated and adverse geological conditions including fractures, spalling and others may appear. However, due to the limitation of testing methods and economic budget, the geological conditions always could not be fully investigated before excavation. In this context, high risk in the aspects of TBM breakdown, project delay and crew causalities is induced. For example, an open TBM used in Jilin Province in China encountered 33 fault and fracture zones while tunneling, which caused serious delay of construction schedule (Parise M, 2008).
