Multi-Objective Physics-Guided Recurrent Neural Networks for Identifying Non-Autonomous Dynamical Systems
This work addresses the trade-off between modeling effort and accuracy in system identification for non-autonomous systems, offering a solution that maintains physical plausibility, though it is incremental as it builds on existing physics-based and data-driven methods.
The paper tackles the problem of modeling non-autonomous dynamical systems by proposing a physics-guided hybrid approach that combines a physics-based model with a recurrent neural network, resulting in substantial accuracy improvements over purely physics-based models on real data.
While trade-offs between modeling effort and model accuracy remain a major concern with system identification, resorting to data-driven methods often leads to a complete disregard for physical plausibility. To address this issue, we propose a physics-guided hybrid approach for modeling non-autonomous systems under control. Starting from a traditional physics-based model, this is extended by a recurrent neural network and trained using a sophisticated multi-objective strategy yielding physically plausible models. While purely data-driven methods fail to produce satisfying results, experiments conducted on real data reveal substantial accuracy improvements by our approach compared to a physics-based model.