In the ever-evolving landscape of maritime operations, a groundbreaking study led by Qiuyu Li from the Merchant Marine College at Shanghai Maritime University has introduced a novel approach to cylinder lubrication that could revolutionize how ships maintain their main engines. Published in the Journal of Marine Science and Engineering, the research presents a lightweight image-based decision support model that leverages deep learning and computer vision to optimize lubrication strategies, particularly in the context of low-sulfur fuel adoption.
The study addresses a critical challenge faced by the maritime industry: the instability and high costs associated with traditional oil injection strategies, which often rely heavily on manual experience. Li and his team propose a model that comprises three key components: a backbone network, a feature enhancement module, and a similarity retrieval module. The backbone network, EfficientNetB0, efficiently extracts features with low computational overhead. MobileViT Blocks are then integrated to combine the local feature perception of Convolutional Neural Networks (CNNs) with the global modeling capacity of Transformers. To further enhance the model’s performance, Receptive Field Blocks (RFB) and the Convolutional Block Attention Module (CBAM) attention mechanism are introduced, improving the receptive field and multi-scale representation, and focusing on salient regions for better feature discrimination.
The model’s effectiveness was tested using a high-quality image dataset constructed from WINNING’s large bulk carriers under various sea conditions. The results are impressive: the EfficientNetB0 + RFB + MobileViT + CBAM model achieved a precision of 99.71%, recall of 99.69%, and an F1-score of 99.70%. These figures represent significant improvements over the baseline EfficientNetB0, with increases of 11.81%, 15.36%, and 13.62%, respectively. Moreover, the model’s computational cost is minimal, with only a 0.3 GFLOP and 8.3 MB increase in model size, striking a balance between accuracy and inference efficiency.
The commercial impacts of this research are substantial. As the maritime industry continues to grapple with the challenges posed by low-sulfur fuel regulations, this model offers a promising solution for optimizing cylinder lubrication. By reducing the reliance on manual experience and improving the stability and cost-effectiveness of oil injection strategies, the model can help ship operators enhance their operational efficiency and reduce maintenance costs.
Li emphasizes the potential of this technology, stating, “Our model demonstrates good robustness and application stability in real-world ship testing, with potential for further adoption in the field of intelligent ship maintenance.” This statement underscores the model’s practical applicability and its potential to drive innovation in the maritime sector.
In conclusion, the research led by Qiuyu Li represents a significant advancement in the field of marine main engine maintenance. By leveraging deep learning and computer vision, the proposed model offers a lightweight, efficient, and highly accurate solution for cylinder lubrication. As the maritime industry continues to evolve, this technology could play a crucial role in enhancing operational efficiency and reducing costs, ultimately contributing to a more sustainable and profitable maritime sector.