Sensitivity of TBM’s Performance to Structural, Control and Geological Parameters Under Different Prediction Models

In general design and analysis of a tunnel boring machine (TBM), many analytical models are proposed to predict the TBM's performance. Various models may result in different performance predictions for the same TBM excavating under the same geological conditions. Therefore, it is essential to perform the quantitative analysis of the impacts from different prediction models and the corresponding key input factors on the TBM's performance. Recently, there is almost no relevant study on such issues for TBM and it is urgent to fill this gap. In this paper, by comparing and analyzing the TBM's performance using different prediction models, three types of total thrust prediction models (the rapid-growth type, the intermediate type, and the slow-growth type) and two types of total torque prediction models (the rapid-growth type and the slow-growth type) are classified and defined for the first time in the TBM-related fields. Then, a global sensitivity analysis (SA) of TBM's performance using the Sobol' method is developed regarding key input factors, including control, structural, and geological parameters. It is found that the relative impacts of the input factors to TBM's performance vary appreciably with the selection of prediction models. Specifically, a global SA on the minimized construction period of a tunneling project with respect to structure parameters is performed. The results show that the structure parameters have similar impacts on the minimized construction period irrespective of the selection of prediction models. The impacts of different prediction models on the minimized construction period of a tunneling project using Genetic Algorithm (GA) are investigated by finding the optimal control and structure parameters. The results interestingly show that the selection of the TBM's performance prediction models has a marginal impact on the minimized construction period but yields partly different key parameters.