Intelligent Prediction Model for Ship Roll Damping Based on Outlier-modified CTGAN
To prevent significant ship rolling in harsh marine environments,an intelligent prediction model for ship rolling damping is proposed based on outlier-modified CTGAN(conditional tabular generative adversarial networks).An RDD dataset was constructed using geometric model data and ship motion data collected from marine observations and simulations.Then CTGAN is employed to generate the rolling damping data by learning the distribution through adversarial training,thereby balancing and augmenting the quantity of the data.An outlier-modified algorithm based on DBSCAN is proposed.In low-density cluster regions,new samples are generated through interpolation using mutual information to reduce the impact of outliers on the model's performance.An ensemble learning algorithm is utilized to build the intelligent prediction model.Experimental results demonstrate that the model achieves significantly better performance than other comparative models with MSE(mean squared error),MAE(mean absolute error),and R2(coefficient of determination)being 0.239,0.449,and 0.982 respectively.The model can provide valuable information for predicting ship rolling motions and contributing to progress in ship engineering.
ship roll dampingintelligent predictiongenerative adversarial networkoutlier modificationensemble learning
NETL
NSTL
万方数据
