An extreme gradient boosting technique to estimate TBM penetration rate and prediction platform

An accurate prediction of the penetration rate (PR) of a tunnel boring machine (TBM) is essential for the schedule and cost estimation of tunnel excavation. To better meet the needs of modern information construction, more computer technologies are being used to integrate the analysis and management of construction data. Herein, an online prediction platform based on a data mining algorithm using ensemble learning (extreme gradient boosting (XGBoost)) is developed for TBM performance prediction. The platform establishes the model and displays the prediction results, while storing a considerable amount of machine data, and providing services for TBMs of multiple projects simultaneously. In establishing the prediction model, users can change the algorithm parameters according to the engineering situation. The prediction capabilities of the platform are demonstrated by 200 field samples obtained from the Songhua River water conveyance project in Jilin. The mean absolute percentage error, coefficient of determination, root mean squared error, variance account for (VAF), and a20-index of the PR are 6.07%, 0.8651, 3.5862, 87.06%, and 0.925, respectively. The results show that the prediction model has a reliable prediction accuracy, which is higher than that of the gradient boosting decision tree, and these results can be displayed on the online platform. It provides effective help for TBM intelligent tunneling.

Automated summary

This article introduces the application of data mining algorithm based on ensemble learning in TBM performance prediction and develops an online prediction platform. The prediction capabilities of the platform are validated by actual samples from the Songhua River water conveyance project in Jilin. The results show that it has reliable prediction accuracy and provides effective help.