Research on Fault Diagnosis Technology for Ship Navigation Systems Based on a Single-Feature Framework and Particle Swarm Optimization-LSTM

To address the challenge that the operational state change laws of key equipment in complex ship systems are difficult to accurately characterize under complex sea conditions and multi-operating modes, this paper proposes an equipment state prediction method based on Single-Feature Independent Modeling (SFIM) and adaptive parameter optimization. From the perspective of equipment-level modeling, the method establishes independent models for the state features of equipment with different physical properties in the navigation task channel. This avoids the negative impact of mutual interference between different features in multi-feature joint modeling on prediction accuracy, enabling the model to focus more on the temporal evolution laws of individual features. On this basis, the Particle Swarm Optimization (PSO) algorithm is introduced to adaptively optimize the key hyperparameters of the Long Short-Term Memory (LSTM) network, ensuring that the model parameter configuration matches the temporal variation characteristics of different equipment state features. Taking representative equipment state features in the navigation task channel, such as vibration, temperature, noise, and deformation, as research objects, prediction modeling and comparative analysis are carried out based on ship model test data. Experimental results show that the proposed PSO-SFIM-LSTM model achieves high prediction accuracy in various feature prediction tasks. Specifically, in temperature feature prediction, its mean squared error (MSE) is reduced by more than an order of magnitude compared with the GRU and CNN-1D models; in vibration and noise feature prediction, the prediction error is reduced by approximately 20%-40%, demonstrating better prediction stability and feature adaptability. Furthermore, the model is applied to the analysis of ship equipment operational data. The results indicate that when the equipment state evolves from stable operation to abnormality, the prediction curve exhibits a trend deviation in advance of the obvious changes in the actual state, providing an effective basis for the early identification of potential equipment faults and operational maintenance decisions.