基于长短期记忆神经网络的船体及螺旋桨性能退化评估

Assessment of Hull and Propeller Performance Degradation Based on Long Short-term Memory Neural Network

采用线性插值和时空插值构建包含气象、航行性能和机舱信息的船舶综合数据库,经统计分析筛选出9个输入变量,基于长短期记忆神经网络建立航速预测模型,获得由预测航速与实测值的残差描述的船体及螺旋桨性能退化的特征参数,建立以航速残差的时间序列为路径的船体及螺旋桨性能退化评估方法。基于某30万吨散货船一年的航行数据进行性能退化评估,航速预测结果的均方误差为0.01,目标船的性能退化路径具有时间相关性和单调变化的趋势。

The integrated database containing marine meteorological data,navigation and engine room information is analyzed by application of statistical method.And nine main variables from the database are selected to develop the speed prediction model by using a long and short-term memory neural network(LSTM).The residuals between the predicted and measured ship speed are applied to describe the characteristics of hull and propeller performance degradation.Based on the time series of speed residual,the degradation path of hull and propeller performance is established.The evaluating method of hull and propeller performance degradation with the ship speed prediction model is validated with one-year voyage data of a 300,000 DWT bulk carrier.The mean square error of the predicted speed is about 0.01,and the performance degradation path of the hull and propeller of the bulk carrier is monotonically decreased with the time.