AI-Driven Approach to Subsea Cable Lay Operability: An LSTM Neural Network for Predicting Cable Dynamic Responses

Abstract This paper presents a novel approach to evaluating the operability of subsea cable lay operations using Artificial Intelligence. Specifically, it introduces a Long Short-Term Memory (LSTM) Artificial Neural Network (ANN) model designed to predict the time-varying dynamics of cable responses—including tension, compression, and curvature—using only vessel motion time-series data. Conventional methods for determining operational weather limits in the offshore cable industry rely on discrete numerical simulations of assumed sea states and simplified weather forecasts, often resulting in overly conservative assessments and reduced operability. Recently, a methodology leveraging real 2D wave spectra and regression analysis to derive equivalent vessel motion limits was proposed to address these limitations, enhancing operability and reliability. Building on these advances, this study develops an LSTM-ANN model that extends the regression-based approach by incorporating multiple parameter relations and capturing complex nonlinear temporal patterns, such as bending hysteresis. The model includes features to improve generalization and prediction accuracy for extreme responses and is trained on a comprehensive OrcaFlex simulation dataset, serving as an efficient surrogate for traditional finite element simulations. Test results demonstrate the model’s high accuracy, particularly in deep-water scenarios where vessel motions predominantly drive cable behavior. This innovative approach improves the precision of cable response predictions during weather window assessments, paving the way for safer and more efficient subsea cable operations.