Development of a Hybrid-Driven Computational Model for Cylinder Pressure in Marine Low-Speed Engines

To address the high operating cost of online cylinder pressure monitoring systems for low-speed engines in ships and the limitations of existing alternatives - i.e., the lack of flexibility of the mechanical model under different operating conditions and the lack of physical interpretability of the data-driven model - this study proposes a hybrid-driven based in-cylinder pressure calculation model. Taking the 6EX340EF marine low-speed engine as the object of study, the method first constructs a mechanical model and optimizes the Wiebe function parameters using the Dung Beetle Optimizer (DBO). Subsequently, the mapping relationships between operating parameters, Wiebe parameters, initial compression stage temperature and charge mass are learned by constructing a combined neural network of Convolutional Neural Network (CNN) and Bi-directional Long and Short-Term Memory Network (Bi-LSTM). Finally, the overall calculation of in-cylinder pressure was realized by integrating a multidimensional parametric framework of engine configuration parameters, real-time running inputs and dynamic MAP maps. The results show IMEP R2 = 0.9864 and peak pressure error ≤ 2%, confirming that the model can provide technical support for long-term real-time pressure measurement and closed-loop optimization control based on in-cylinder pressure for marine low-speed engines.