Genetic algorithm-assisted an improved AdaBoost double-layer for oil temperature prediction of TBM

The oil is one of the main power sources of Tunnel Boring Machine (TBM). In practical engineering, it is necessary to keep the oil temperature within the normal range because too high oil temperature would increase the probability of hitch. Notably, the oil temperature of TBM is affected by numerous factors, which is difficult to predict. To address this issue, genetic algorithm (GA)-assisted an improved AdaBoost double-layer learner (GA-ADA-RF) is proposed. In the GA-ADA-RF, random forests as weak learners of AdaBoost and the method of random sampling with replacement is introduced to construct a double-level learner. In the process of hyperparameter adjustment, the maximum number of trees and the maximum depth of trees are selected as the design variables and the fitness function is established by using the classification evaluation of the algorithm. Compared with the other state-of-the art algorithms, the GA-ADA-RF has a better prediction performance. i.e., Accuracy = 0.985,AUC = 0.991. The GA-ADA-RF also has served in other complex projects similar to TBM and shows potential.

Automated summary

In this study, a method based on genetic algorithm and improved AdaBoost double-layer learner (GA-ADA-RF) is proposed to predict the oil temperature of TBM. The GA-ADA-RF constructs a double-level learner by introducing random forests as weak learners of AdaBoost and using random sampling with replacement. The fitness function is established using the classification evaluation of the algorithm. Compared to other algorithms, the GA-ADA-RF has better prediction performance and has shown potential in complex projects similar to TBM.