Cross-project prediction for rock mass using shuffled TBM big dataset and knowledge-based machine learning methods

Extensive research has confirmed the successful prediction of rock mass quality in tunnel boring machine(TBM)construction using machine learning methods based on big data collected during the boring process.However,the developed model cannot be applied to a new project owing to the different mechanical and environmental features involved in different projects.This study tries to combine the datasets of three TBM projects whose cutterhead diameters are 5.2,7.9 and 9.8 m,respectively.In this study,machine learning focused on predictions of a binary rock mass quality system was implemented using this unified dataset by adding the diameter and disc cutter number as new attributes into the input.The process consists of:(1)individual learning for the three respective datasets,(2)shuffled learning for the unified dataset containing randomly distributed information from the three projects,and(3)crossed learning aimed at validating that the algorithm developed on the unified dataset can produce predictions with equally acceptable accuracies as those obtained in the individual learning.It is anticipated that with more datasets joining this cross-project learning,we will be able to develop a machine learning algorithm that is suitable for new projects with a wide range of cutterhead diameters and disc cutter numbers at the beginning of the tunnel excavation.

THE PROJECTIVE PLANE CROSSING NUMBERS OF CIRCULAR GRAPHS

The authors give an upper bound for the projective plane crossing number of a circular graph. Also, the authors prove the projective plane crossing numbers of circular graph C （8, 3） and C （9, 3） are 2 and 1, respectively.