| 摘要:基于深度学习，以6L34DF型船用双燃料柴油机实船数据构建多维时间序列为源，在双向门控循环单元（BiGRU）基础上结合相关性分析与主成分分析（PCA）技术，深度提取特征、去除冗余、降维并缩时。构建 PCA-BiGRU预测模型与BiGRU对比，结果表明PCA-BiGRU模型在平均绝对误差、均方误差及均方根误差上分别降低0.106、0.204和0.148，精度大幅提升。该研究为船舶柴油机状态监测预测及其他智能化船舶设备运行状态预测提供新思路，融合PCA与BiGRU优化模型性能，有力推动船舶智能化监测技术发展。 |
| 关键词: 船舶柴油机 排气温度 主成分分析 双向门控循环单元 |
|
| Prediction of Exhaust Temperature Trend of Marine Diesel Engine Based on Deep Learning |
| LIU Ben,LIU Yefei,LIU Wenke,MA Lisheng |
| Jiangsu Maritime Institute ， Nanjing 211100，China;Nanjing Science and Technology Achievement Transfommation Service Center， Nanjing 211100，China |
| Abstract:Based on deep learning，the real-ship data of the 6L34DF marine dual-fuel diesel engine was used to construct multi-dimensional time series as the data source.Based on the bidirectional gated recurrent unit （BiGRU），combined with Correlation Analysis and Principal Component Analysis （PCA），the features were deeply extracted，redundancies were removed，dimensions were reduced，and computing time was shortened. A PCA-BiGRU prediction model was constructed and compared with the BiGRU model.The results show that the PCA-BiGRU model reduces the mean absolute error，mean square error and root mean square error by 0.106， 0.204 and 0.148， respectively，with a significant improvement in accuracy.This research provides new ideas for the condition monitoring and prediction of marine diesel engines and the prediction of the operating state of other intelligent ship equipments.The integration of PCA and BiGRU optimizes the model performance and vigorously promotes the development of ship intelligent monitoring technology. |
| Key words: marine diesel engine exhaust temperature Principal Component Analysis bidirectional gated recurrent unit |
