Fig 1.
Damage to wind turbine foundation, (a) Anchor rod fracture situation, (b) Surface cracks of wind turbine foundation.
Fig 2.
Schematic diagram of wind turbine reinforcement, (a) Wind turbine reinforcement on-site construction, (b) Wind turbine reinforcement diagram.
Fig 3.
Layout of monitoring points.
Fig 4.
Partial data variation graph of monitoring, (a) Anchor cable stress variation, (b) Steel bar stress variation, (c) Concrete stress variation.
Fig 5.
Illustrates the effectiveness of anomaly detection for features S1-4 across various algorithms, (a) Iterative Rolling Difference-Z-score, (b) Isolation Forest, (c) One-Class SVM, (d) DBSCAN, (e) LOF, (f) K-Means, (g) Gaussian Mixture Model.
Fig 6.
Detection of large-scale missing values and outliers.
Fig 7.
Visualization of missing data.
Fig 8.
Data imputation framework.
Table 1.
Performance of different algorithms on the test set.
Table 2.
Parameters corresponding to different algorithms.
Fig 9.
Comparison chart of linear interpolation repair for individual missing values.
Table 3.
Evaluation of different missing data scenarios.
Fig 10.
Residual plots under different conditions (a) 60 missing data points (b) 120 missing data points (c) 200 missing data points.
Fig 11.
Phase 2 model parameter test plot (a) The reduction of outliers with iterations, (b) Relationship between K value and MSE.
Fig 12.
Comparison of imputed data with original data.
Fig 13.
Kernel density plots of imputed data vs. Original data (a)51% Data Missing (C2-1), (b) 80% Data Missing (S1-2), (c) 38% Data Missing (D5-2), (d) 33% Data Missing (MS-3).
Fig 14.
Comparison of mean and variance between imputed data and original data.
Fig 15.
Continued.