Shield attitude prediction based on Bayesian-LGBM machine learning

Abstract

Effective shield attitude control is essential for the quality and safety of shield construction. The traditional shield attitude control method is manual control based on a driver's experience, which has the defects of hysteresis and poor reliability. This research proposes an intelligent method to predict the shield attitude based on a Bayesian-light gradient boosting machine (LGBM) model. The constructed model includes 29 parameters that impact the shield attitude and 6 parameters that represent the shield attitude. The developed the Bayesian-LGBM model can predict the shield attitude and support shield attitude control by adjusting construction parameters and conducting iterative prediction. Guiyang rail transit line 3 is selected as a case study to verify the effectiveness of the proposed method. The results indicate that: (1) The developed Bayesian-LGBM model is able to effectively predict the shield attitude; (2) The importance ranking can clarify the key construction parameters that should be controlled; (3) The proposed method enables supporting the effective shield attitude control by continuously adjusting the shield construction parameters. The proposed attitude guidance control method based on the proposed Bayesian-LGBM model can be used to provide a reference for actual shield attitude applications and other similar problems.