The invention discloses a numerical sample and
random forest-based TBM jamming risk prediction method. The method comprises the steps: building a refined numerical
simulation model, simulating the aging deformation of surrounding rock based on a
creep damage model, and achieving the
simulation of a TBM construction process; setting values of different card
machine influence factors in the numerical
simulation model, and calculating numerical samples containing different working conditions; establishing a jamming risk discrimination index, and calibrating the jamming
risk level of the sample; and establishing a
random forest model, performing model training based on the numerical samples, and predicting the jamming
risk level of the actual construction section by using the trained
random forest model. According to the method, the jamming numerical value sample
library is constructed based on refined numerical value simulation, and the problems of few monitoring samples and imbalance existing in application of
machine learning in
engineering are solved; by using the trained random forest model, the jamming
risk level of the actual construction section can be quickly predicted, so that early prevention and control of disasters are guided, and safe and efficient construction of the TBM is ensured.