Research on a Real-Time motion forecasting model for marine cargo elevators based on EMD-KPCA-LSTM

Currently, traditional single Long Short-Term Memory(LSTM) models exhibit reduced prediction accuracy and increased errors when handling nonlinear data influenced by complex maritime environments, particularly during periods of short-term, intense fluctuations in data points. To address these challenges, a novel EMD-KPCA-LSTM composite model is proposed. This model integrates Empirical Mode Decomposition (EMD) and Kernel Principal Component Analysis (KPCA) to optimize LSTM performance, enabling real-time motion prediction for marine cargo elevators during operational navigation conditions. Simulation results indicate that compared to the single LSTM model, the composite model exhibits significantly reduced prediction errors relative to actual values. Based on the Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) evaluation metrics, the errors decreased by 29.3% and 38.2%, respectively. The EMD-KPCA-LSTM model significantly enhances the prediction accuracy and stability of time-series data in complex marine environments, effectively overcoming the sensitivity of single LSTM model to short-term sharp fluctuations. This model provides a reliable solution for processing high-noise, nonlinear marine environmental data and holds significant engineering application value in the field of marine equipment condition monitoring and fault early warning.