Comprehensive evaluation of machine learning algorithms applied to TBM performance prediction

To date, the accurate prediction of tunnel boring machine (TBM) performance remains a considerable challenge owing to complex interactions between TBM and ground. Using evolutionary polynomial regression (EPR) random forest (RF), this study develops two novel models for performance. Both can predict penetration rate field index as outputs with four input parameters: uniaxial compressive strength, intact rock brittleness index, distance planes weakness, angle axis weakness (α). First, performances both EPR- RF-based are examined by comparison conventional numerical method (i.e., multivariate linear regression). Subsequently, RF- EPR-based further investigated compared, including model robustness unknown datasets, interior relationships output parameters, variable importance. The results indicate that has greater accuracy, particularly in identifying outliers, whereas is more convenient use engineers its explicit expression. accurately identify parameters. This ensures their excellent generalization ability high accuracy on datasets.