Deep Learning-Based Method for Assessing the Performance Degradation of Marine Diesel Engines

To overcome the challenge of quantifying performance degradation in marine diesel engine operation monitoring, a hybrid model is proposed. A one-dimensional convolutional neural network (1D-CNN) module is used to extract local spatiotemporal features, and a long short-term memory network (LSTM) is employed to learn dependencies. The final result uses power attenuation rate as the performance degradation index, and the connections are made through fully connected layers. Experiments based on 12 months of actual operating data from a MAN B&W 6S35MC diesel engine validate the effectiveness of the proposed method. The results show that the mean absolute error of the relative power attenuation rate is 1.68%, and the mean trend consistency is 0.981. The proposed method can effectively quantify the performance degradation degree of diesel engines and can be effectively applied to intelligent maintenance.