原油船海上航行升沉运动Bayes-LSTM预测方法

Prediction of crude oil tanker heave motion based on Bayes-LSTM

GAO Liangjun; TANG Yixin; CHEN Liang; WNANG Beifu

升沉运动; STAR CCM+; 贝叶斯算法; Bayes-LSTM; LSTM

Heave motion;  ; STAR CCM+;  Bayesian algorithm;  Bayes-LSTM;  LSTM

为了更好预测船舶在海上航行中的升沉运动，提高船舶海上航行与作业安全水平，以10×10⁴ t级原油船为研究对象，利用船舶模型运动过程数值模拟软件STAR CCM+构建其仿真模型，由无液货舱和半载液货舱两种情况及0.5λ、1.0λ、1.5λ（λ=6.16 m）3种波长组合构成6种工况，获取6组升沉运动数据，并将其以8:2的比例划分为训练集和测试集，利用贝叶斯算法优化后的长短期记忆神经网络（Bayes-LSTM）模型进行模型升沉运动预测，将预测结果与长短期记忆神经网络（LSTM）模型的预测结果进行对比。结果表明：Bayes-LSTM模型比LSTM模型的预测精度最大提高3倍以上，显示出Bayes-LSTM模型对船舶海上航行和作业过程中升沉运动预测的优势。

The research on the prediction of ship heave motion is beneficial to the safe navigation and operation of ships at sea. Based on STAR CCM+, this paper takes a 100,000 ton crude oil tanker as a simulation model. Six sets of heave motion data are obtained under three wavelengths from two angles of free cargo tank and half-loaded cargo tank. The 6 working conditions data obtained were predicted and analyzed by using the Bayes-LSTM neural network model optimized by Bayes algorithm, and the prediction results of the LSTM neural network model were compared. The results show that Bayes-LSTM model can improve the maximum prediction accuracy by three times compared with LSTM model. Thus it can be concluded that Bayes-LSTM model is a reliable method in the research of ship heave motion prediction.