Cross Project Conversion Relationship of Key Parameters of TBM Rock Breaking
A large amount of data have been collected in TBM information construction,and the establishment of machine learning model through data mining is the premise of realizing TBM intelligence.However,at the initial stage of TBM construction,the prediction performance of machine learning model is poor due to the lack of data;At the same time,due to the differences in TBM equipment structure and cutterhead diameter,the machine learning model based on historical projects training is not suitable for new projects.In order to solve this bottleneck problem,the physical invariants only related to the number of cutters and the diameter of the cutter head are derived from the force analysis of a single cutter,the empirical method and the torsional shear experimental model.The new projects data can be converted by using the conversion scheme composed of invariants;Then,the conversion scheme of key parameters of rock breaking with the best application performance in surrounding rock classification and machine learning model is selected;Then,the genetic algorithm is used to iteratively search the optimal conversion scheme invariant which is suitable for the current project.The research results show that the data of Yinchao project(new project)are input into the machine learning model of Yinsong project(historical project)after"invariant"conversion,and the prediction performance R2 of cutterhead torque T and cutterhead thrust F reach 0.84 and 0.70 respectively.By using this conversion scheme invariant,the TBM construction data of different projects can be normalized and analyzed under the same framework,and the machine learning model trained based on historical project data is realized to guide the construction of new projects.The research results can provide reference for the cross project application of TBM machine learning model.
TBMmachine learningkey parameters of rock breakinginvariantgenetic algorithmgeotechnical engineeringengineering geology
NETL
NSTL
万方数据
