.png){kind=icon}
.png){kind=icon}
Return
Deep learning-enhanced simulation-based inference for improving digital twins of marine vessels
.png){kind=icon}
OA
AI
S
A
DOI：10.1016/j.apor.2026.105084.png)
.png){kind=icon}
.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
This paper presents a novel framework, Deep Learning-Enhanced Simulation-Based Inference (SBI), to improve the accuracy of digital twins for marine vessels. In the proposed framework, deep learning is integrated into the SBI process to effectively extract informative representations from both sequential and static data. To this end, a Long Short-Term Memory (LSTM) network is employed to process time-series data generated from vessel simulations, capturing temporal dependencies and complex dynamic patterns. The LSTM-generated embeddings are combined with static statistical features through a hybrid embedding module, resulting in a comprehensive and robust data representation. This fused representation is then utilized by a neural density estimator within the Neural Posterior Estimation (NPE) paradigm to approximate the posterior distribution of model parameters with high accuracy. The proposed method is evaluated on simulated datasets representative of marine vessel dynamics. The results demonstrate that incorporating deep learning into SBI significantly enhances the predictive performance and accuracy of digital twins compared to conventional approaches. By leveraging the strength of deep neural embeddings and Bayesian inference, the framework provides a more robust and data-driven solution for parameter estimation in marine vessel modeling.
Keywords:
Simulation-based inference
Deep learning
Marine vessel
Digital twins
AI Summary
Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.
Journal
IF:
4.4
Papers:
4.0K
Citations:
1.3W
Organization
No organization information available
Cited Papers