Research on Intelligent Prediction for Deep and Large-Diameter Circular Shaft in the Pearl River Delta Region

In order to realize the complementary resources, a series of large-scale water resource allocation projects have been carried out in China, involving fussy designs of shaft structures under complex stratigraphic conditions. In Pearl River Delta Water Resources Allocation Project, for example, 28 circular shield tunneling shafts are arranged, with depths ranging from 34 to 74 m and diameters ranging from 24 to 39 m. Unlike conventional shield tunneling shafts, mechanical performance for the deep and large-diameter circular shafts presents a significant spatial effect, resulting in fussy structural design and inefficiency. In this paper, deformation measurement data of the shafts are first collected and systematically analyzed. A 3D finite element (FE) model is established to consider spatial effects and non-uniform distribution of the adjacent stratum. Furthermore, we accomplished a series of batch processing tasks, including FE simulations under different conditions and extraction of safety control indicators. As a result, we have established a database composed of geometric dimensions, formation parameters, and deformation control indicators. Finally, using machine-learning methods such as Random Forest (RF) and XGBoost for training, a deformation prediction system for the circular shafts in the Pearl River Delta region is constructed based on intelligent information processing technology. The results are quantitatively consistent with the measurement data and FE simulation value, providing an important basis for shaft positioning and structural optimization at the initial stage of design.