A novel gradient boosting regression tree technique optimized by improved sparrow search algorithm for predicting TBM penetration rate

In tunnel mechanized excavation, the tunnel boring machine (TBM), as fast, reliable, and environment-friendly equipment, has been widely used in tunnel engineering. Accordingly, it is of great significance to understand the performance of TBM and minimize the related risks in the operation process of TBM. The purpose of this research is to propose a new hybrid model for TBM performance prediction (like penetration rate, PR) based on the improved sparrow search algorithm-gradient boosting regression tree technique (ISSA-GBRT). Improving the accuracy of GBRT by improving the SSA algorithm based on chaotic sine map and Student t-distribution mutation is the main advantage of this model. For comparison, other hybrid models based on GBRT were applied and proposed, namely, particle swarm optimization (PSO)-GBRT, biogeography-based optimization (BBO)-GBRT, and sparrow search algorithm (SSA)-GBRT. To do so, a Chinese case study, the Shenzhen Metro Line 10 tunnel, was used to validate the developed models, and 80% of the total samples were randomly chosen as the training set, while the remaining 20% of the samples were used to test the models. To compare the results calculated from the model, two evaluation indicators, correlation coefficients (R) and root mean square error (RMSE), were computed. The R and RMSE values of (0.871, 0.120), (0.856, 0.129), (0.858, 0.131), and (0.852, 0.133) were obtained for testing datasets of ISSA-GBRT, SSA-GBRT, PSO-GBRT, and BBO-GBRT, respectively, which confirmed that the developed ISSA-GBRT was an accurate model for the prediction of TBM PR. In addition, the feature importance analysis found that the impacts of total thrust, cutter torque, cutter speed, and the permeable coefficient on PR are more significant than other factors.