A Ship Equipment Suitability Assessment Method Based on Multi-Scale CNN and LSTM Model

To address the issues that traditional ship equipment applicability assessment methods rely on manual feature extraction and have insufficient generalization ability in complex maritime environments, This research introduces a hybrid deep learning architecture that integrates multi-scale convolutional layers with a bidirectional Long Short-Term Memory (LSTM) network augmented by an attention mechanism. The model is designed to directly map the operational status of key components to the overall applicability of the equipment, forming an end-to-end assessment framework. This study takes the operating status of key mechanical components of the ship’s power system as the core basis for evaluating the applicability of the entire equipment. The model adopts a multi-branch one-dimensional CNN structure to automatically extract multi-scale features from equipment operation data, enhances the ability to focus on key evaluation indicators by introducing an attention mechanism, and uses stacked bidirectional LSTM to mine the temporal dependencies of data. Experimental results show that the average assessment accuracy of the proposed method under 7 applicability levels reaches 99.54%, which is significantly improved compared with traditional assessment methods. It can still maintain high accuracy in noisy environments, demonstrating the effectiveness and robustness of the method in complex maritime scenarios.