Abstract The full-section hard rock tunnel boring machine (TBM) is affected by different factors during tunneling, such as deadweight, geological conditions and human factors, resulting in a change in the tunneling posture and a deviation from the target axis, so derivation correction and direction adjustment to the target trajectory are required to ensure the quality of construction. A DTL-based TBM derivation correction and direction control method was put forward in this paper, i.e. the DTL neural network was used to build a parameter prediction model for TBM derivation correction and direction control, and then the TBM derivation correction and direction control position and posture model were analyzed to plan the derivation correction trajectory in combination with the maximum amount of movement of edge cutter and the minimum turning radius. The engineering verification results showed that the DTL-based parameter prediction model for TBM derivation correction and direction control model had a higher control accuracy which could confine the difference between TBM posture and target axis within ±20 mm, and the surface of tunnel walls was smoother, improving the quality of tunnel construction; the derivation correction trajectory was planned based on the maximum amount of movement of edge cutter and the minimum turning radius, and the derivation correction process was more controllable to avoid cutters and cutterhead from being damaged due to over-ad?justment and also reduce the risks of TBM jamming.
Abstract：
The full-section hard rock tunnel boring machine (TBM) is affected by different factors during tunneling, such as deadweight, geological conditions and human factors, resulting in a change in the tunneling posture and a deviation from the target axis, so derivation correction and direction adjustment to the target trajectory are required to ensure the quality of construction. A DTL-based TBM derivation correction and direction control method was put forward in this paper, i.e. the DTL neural network was used to build a parameter prediction model for TBM derivation correction and direction control, and then the TBM derivation correction and direction control position and posture model were analyzed to plan the derivation correction trajectory in combination with the maximum amount of movement of edge cutter and the minimum turning radius. The engineering verification results showed that the DTL-based parameter prediction model for TBM derivation correction and direction control model had a higher control accuracy which could confine the difference between TBM posture and target axis within ±20 mm, and the surface of tunnel walls was smoother, improving the quality of tunnel construction; the derivation correction trajectory was planned based on the maximum amount of movement of edge cutter and the minimum turning radius, and the derivation correction process was more controllable to avoid cutters and cutterhead from being damaged due to over-ad?justment and also reduce the risks of TBM jamming.
HOU Kunzhou
.Study on TBM Deviation Correction and Direction Control Based on the Deep Transfer Learning (DTL)[J] MODERN TUNNELLING TECHNOLOGY, 2022,V59(4): 81-89
2022, Vol. 59 