Learning-based model augmentation with LFRs
Fuente:
arXiv
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866908495768453120 |
|---|---|
| author | Hoekstra, Jan H. Verhoek, Chris Tóth, Roland Schoukens, Maarten |
| author_facet | Hoekstra, Jan H. Verhoek, Chris Tóth, Roland Schoukens, Maarten |
| contents | Nonlinear system identification (NL-SI) has proven to be effective in obtaining accurate models for highly complex systems. In particular, recent encoder-based methods for artificial neural networks state-space (ANN-SS) models have achieved state-of-the-art performance on various benchmarks, while offering consistency and computational efficiency. Inclusion of prior knowledge of the system can be exploited to increase (i) estimation speed, (ii) accuracy, and (iii) interpretability of the resulting models. This paper proposes an encoder-based model augmentation method that incorporates prior knowledge from first-principles (FP) models. We introduce a novel \linear-fractional-representation (LFR) model structure that allows for the unified representation of various augmentation structures including the ones that are commonly used in the literature, and an identification algorithm for estimating the proposed structure together with appropriate initialization methods. The performance and generalization capabilities of the proposed method are demonstrated in a hardening mass-spring-damper simulation. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2404_01901 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Learning-based model augmentation with LFRs Hoekstra, Jan H. Verhoek, Chris Tóth, Roland Schoukens, Maarten Systems and Control Nonlinear system identification (NL-SI) has proven to be effective in obtaining accurate models for highly complex systems. In particular, recent encoder-based methods for artificial neural networks state-space (ANN-SS) models have achieved state-of-the-art performance on various benchmarks, while offering consistency and computational efficiency. Inclusion of prior knowledge of the system can be exploited to increase (i) estimation speed, (ii) accuracy, and (iii) interpretability of the resulting models. This paper proposes an encoder-based model augmentation method that incorporates prior knowledge from first-principles (FP) models. We introduce a novel \linear-fractional-representation (LFR) model structure that allows for the unified representation of various augmentation structures including the ones that are commonly used in the literature, and an identification algorithm for estimating the proposed structure together with appropriate initialization methods. The performance and generalization capabilities of the proposed method are demonstrated in a hardening mass-spring-damper simulation. |
| title | Learning-based model augmentation with LFRs |
| topic | Systems and Control |
| url | https://arxiv.org/abs/2404.01901 |