Metraipolis › archive › dc_674292972129

Physics-Guided Inferencer-in-the-Loop Digital-Twin for Real-Time Adaptive Emulation of Nonlinear PMSMs — arXiv:2610.38149

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.38149
License (as archived)
CC0-1.0
Archived
2026-11-01T02:59:28.554Z
SHA-256 of the archived text
0713e65025d40d1bc254265e7abcb358cea9ae99729b816db5ce80e8d7f6e223
Stable URL of this page
https://metraipolis.com/documento/dc_674292972129

How to cite

Physics-Guided Inferencer-in-the-Loop Digital-Twin for Real-Time Adaptive Emulation of Nonlinear PMSMs — arXiv:2610.38149 — arXiv (en), https://arxiv.org/abs/2610.38149 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_674292972129 (Metraipolis archive dc_674292972129, sha256 0713e65025d40d1bc254265e7abcb358cea9ae99729b816db5ce80e8d7f6e223)
BibTeX
@misc{dc674292972129,
  title = {Physics-Guided Inferencer-in-the-Loop Digital-Twin for Real-Time Adaptive Emulation of Nonlinear PMSMs — arXiv:2610.38149},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.38149},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_674292972129; sha256 0713e65025d40d1bc254265e7abcb358cea9ae99729b816db5ce80e8d7f6e223},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.38149. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

A real-time digital-twin (RTDT) of a permanent magnet synchronous machine (PMSM) allows drive controllers to be tested in hardware-in-the-loop (HIL) before they meet the real machine. Its machine model rests on identified parameters or flux look-up tables (LUTs) from finite-element analysis (FEA) or characterization tests; the maps are regenerated offline when the machine changes, and online estimators track a few parameters such as resistance and PM flux. This article makes the executed model learnable from reference data. A physics-guided machine learning (PGML) network advances the currents and predicts torque in one step, trained with analytical one-step labels. In the proposed Inferencer-in-the-Loop (IIL) architecture, a GPU retrains the network while a field-programmable gate array (FPGA) executes the last accepted weights. For damper circuits, shifted parameters, and an FEA-calibrated saturated machine, one kernel form reaches average normalized mean absolute errors (NMAEs) of 0.19%-0.67% at measured FPGA kernel latencies of 0.30-0.34 $\mu$s, and retraining for a shifted machine nears its final loss within 94 s. On a magnetically linear 200-W drive, a residual form corrects the analytical step and learns its physical parameters from operating data; each step takes under 2.5 $\mu$s, 20 times faster than real time and ahead of the physical drive's measurements even at 20-kHz sampling. On six held-out runs (1193 s), its $i_q$ error is 5.9%-9.4% of measured RMS, against 10.4%-13.2% for an independently fitted grey-box model, and over 30 min of loaded heating its online-updated heads track a 4.1% winding-resistance rise with an RMS error of 0.44 m$\Omega$.

Tema degli esploratori: calcolo e infrastrutture (elenco arXiv eess.SY, nuovi lavori del giorno).
Categorie arXiv: Systems and Control (eess.SY).
Pagina del lavoro: https://arxiv.org/abs/2610.38149 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-11-01T02:25:30.724Z dalla pagina https://arxiv.org/list/eess.SY/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).