Fig 1.
CNN-BiLSTM-Atintion model structure.
Table 1.
Parameter setting and robustness optimization strategy of traning.
Fig 2.
The ‘GuoHaiShi 1’ offshore test platform.
Table 2.
The calculation results of different recurrence periods of wind and waves.
Table 3.
Main dimensional parameters and hydrostatic parameters of the platform.
Fig 3.
Finite element model of the “Guohaishi 1” offshore test platform.
Table 4.
IMFs components of ceemdan decomposition of roll.
Table 5.
IMFs components of ceemdan decomposition of heave.
Fig 4.
Ceemdan decomposition of the roll dataset.
Fig 5.
Ceemdan decomposition of the heave dataset.
Fig 6.
Time history curves of roll motion in different models.
Fig 7.
Regression curves of roll motions in different models.
Fig 8.
The per-epoch training loss curves of the roll motion for different models.
Fig 9.
Time history curves of heave motion in different models.
Fig 10.
Regression curves of actual and predicted heave motions in different models.
Fig 11.
The per-epoch training loss curves of the roll motion.
Table 6.
Comparison of evaluation indicators of different models in roll motion.
Table 7.
Comparison of evaluation indicators of different models in heave motion.
Fig 12.
Feature importance score and the upper part is roll motion and the lower part is heave motion.
Table 8.
Physical model scaling solution.
Fig 13.
Scaled model of the floating offshore test platform.
Fig 14.
Wave height sensor.
Fig 15.
TSL Thermal Anemometer.
Fig 16.
Non-contact six-degree-of-freedom acquisition system.
Fig 17.
Schematic diagram of test layout.
Table 9.
Test conditions.
Fig 18.
Field test arrangement of the floating platform scaling model.
Fig 19.
On-site test diagram of data acquisition computing system.
Fig 20.
Regression curve and the per-epoch training loss curves.
Fig 21.
Two-stage EWMA control line and RMSE distribution of the test set.
Fig 22.
Importance Score of each feature and each sample.
Fig 23.
Time series curve of true value and predicted value.
Fig 24.
Cumulative sum graph (Left) and Frequency distribution histogram (Right).
Fig 25.
Anomaly distribution map.