In modern intelligent shipping, ensuring the stable and reliable technical condition of marine diesel engines is critical for safe and efficient vessel operations. Conventional fault diagnosis approaches and many existing Transformer-based methods often focus on single-scale features, potentially overlooking subtle fault indicators and reducing diagnostic accuracy under complex working conditions. To address these limitations, this paper proposes a Multi-Scale Attention Transformer (MSAT) model that integrates both high- and low-resolution attention mechanisms. This multi-scale strategy enhances the extraction of detailed and coarse-grained features, improving the model’s capacity to detect and characterize complex diesel engine faults. Additionally, an optimized Nadam optimizer is employed to refine convergence speed and accuracy, surpassing the Adam-based baseline by 0.71%. Rigorous testing on a publicly available diesel engine fault dataset demonstrates that the MSAT model achieves a diagnostic accuracy of 99.86% at a 60 dB signal-to-noise ratio (SNR), outperforming established models such as GRU and LSTM by more than 1%. Even under severe noise interference (0 dB SNR), the model maintains a high accuracy of 96.86%, highlighting its robustness and suitability for real-time monitoring in challenging marine environments. By quantitatively validating these improvements in diagnostic accuracy and noise resistance, this work offers a novel and effective solution for predictive maintenance and operational condition assessment of marine diesel engines, contributing to the reliability and safety of intelligent shipping systems.

Marine diesel engines are the preferred power equipment for ships and are the most important component among the numerous electromechanical devices on board. Accidents involving these engines can potentially cause immeasurable damage to the vessel, making fault detection in marine diesel engines crucial. This design enables the detection and reporting of faults in marine diesel engines at the earliest possible time through the computation of convolutional neural networks, which is of great significance for ensuring the safe navigation of ships. For this functionality, the Xilinx ZYNQ-7000 XC7Z010 is selected as the main control chip, and the LoRa wireless network is used as the transmission module. The FreeRTOS embedded operating system is ported, with sensor data collection completed on the PS side of the ZYNQ chip and algorithm acceleration calculations on the PL side. Data are then transmitted to the host computer via the LoRa module paired with a custom protocol. Experimental test results show that the program provides stable data transmission, with each module of the algorithm generally accelerating by more than 95% and an accuracy rate of 92.86%. Additionally, the host computer can display the received data in real time. The custom protocol’s header also allows for precise judgments about the completeness and origin of messages, facilitating the expansion of other SOC’s message uplink and the host computer’s message downlink.