Machine Learning-Based Methods for Filling the Missing Values of Large Shield Segment Uplift

The monitoring environment for segment uplift during tunnel construction is highly complex, often resulting in significant data loss. This issue adversely impacts research on the stress development of segment structures and related studies. Based on the Jinan Yellow River tunnel project, a method to fill the missing value of segment uplift was established based on machine learning. R-reliefF algorithm and principal component analysis method were used to carry out feature engineering on the dataset composed of measured segment uplift, shield machine driving parameters, formation parameters, etc. Random Forest algorithm and XGBoost algorithm were used to carry out machine learning training and prediction on the processed dataset, respectively. The missing value filling method is established based on the two machine learning methods above. The results show that: The proposed algorithm could fill the missing value in each stage of the segment uplift. Compared with the Random Forest algorithm, the missing value filling method based on the XGBoost algorithm is more accurate. At the same time, even if the measured missing data reaches 40%, the data filled by this method is still close to reality, and the filling effect is good. The results of this project have important reference values for the measurement of segment uplift and related research of segments in practical engineering.