RHINO-MAG: Recursive H-Field Inference based on Observed Magnetic Flux under Dynamic Excitation
arXiv:2603.29745v1 Announce Type: cross Abstract: Driven by the MagNet Challenge 2025 (MC2), increased research interest is directed towards modeling transient magnetic fields within ferrite material. An accurate time-resolved and temperature-aware H-field prediction is essential for optimizing magnetic components in applications with quasi-stationary / non-stationary excitation waveforms. Within the scope of this investigation, a selection of model structures with varying degrees of physically motivated structure are compared. Based on a Pareto investigation, a rather black-box gated recurrent unit (GRU) model structure with a graceful initialization setup is found to offer the most attractive model size vs. model accuracy trade-off, while the physics-inspired models performed worse. For
View PDF HTML (experimental)
Abstract:Driven by the MagNet Challenge 2025 (MC2), increased research interest is directed towards modeling transient magnetic fields within ferrite material. An accurate time-resolved and temperature-aware H-field prediction is essential for optimizing magnetic components in applications with quasi-stationary / non-stationary excitation waveforms. Within the scope of this investigation, a selection of model structures with varying degrees of physically motivated structure are compared. Based on a Pareto investigation, a rather black-box gated recurrent unit (GRU) model structure with a graceful initialization setup is found to offer the most attractive model size vs. model accuracy trade-off, while the physics-inspired models performed worse. For a GRU-based model with only 325 parameters, a sequence relative error of 8.02 % and a normalized energy relative error of 1.07 % averaged across five different materials are achieved on unseen test data. With this excellent parameter efficiency, the proposed model won the first place in the performance category of the MC2.
Subjects:
Systems and Control (eess.SY); Materials Science (cond-mat.mtrl-sci); Signal Processing (eess.SP)
Cite as: arXiv:2603.29745 [eess.SY]
(or arXiv:2603.29745v1 [eess.SY] for this version)
https://doi.org/10.48550/arXiv.2603.29745
arXiv-issued DOI via DataCite (pending registration)
Submission history
From: Oliver Wallscheid [view email] [v1] Tue, 31 Mar 2026 13:43:52 UTC (1,698 KB)
Sign in to highlight and annotate this article
Conversation starters
Daily AI Digest
Get the top 5 AI stories delivered to your inbox every morning.
More about
modelannounceapplication
How AI is Transforming Demand Planning in Modern Supply Chains (From Forecasting to Execution)
Traditional demand planning struggles with dynamic markets, relying too heavily on historical data. AI changes this by enabling real-time demand sensing, probabilistic forecasting, and reinforcement learning for smarter decisions. By combining data, predictive models, and digital twins, businesses can optimize inventory, reduce stockouts, and respond faster to disruptions—turning planning into a continuous, adaptive system. Read All
The Periodic Table of AI Architecture: Assigning Clear Roles to Scattered AI Findings
A speculative but highly insightful conceptual framework for AI architecture A Mini Textbook for AI Engineers on Structure, Flow, Trace, and Residual Governance.pdf just released on Open Science Framework for public review. This mini-textbook, with detail tutorial notes, offers a unified lens for thinking about intelligent systems — moving beyond “just scale more” toward structured coordination under real limits . It treats advanced AI not as an all-knowing predictor, but as bounded observers that extract stable structure from noisy reality while leaving a governable residual (ambiguity, fragility, and unresolved parts). At its core is a clean grammar built around: Maintained Structure vs. Active Flow Adjudication (separating the viable from the merely possible) Semantic time (event-define
Knowledge Map
Connected Articles — Knowledge Graph
This article is connected to other articles through shared AI topics and tags.
More in Products
Governing the Integration Layer: How Real Time Data Platforms Accelerate M&A Execution at Scale
Most mergers fail not because of strategy, but due to poor technology integration. Without a structured integration layer, systems remain fragmented, data becomes inconsistent, and decision-making slows. A well-designed integration architecture combined with real-time data creates a unified view of the business, enabling faster decisions, operational efficiency, and quicker realization of synergies. Read All
A Chinese robotics startup with a Tesla Optimus rival is seeking a new chief scientist with an $18 million salary - Business Insider
A Chinese robotics startup with a Tesla Optimus rival is seeking a new chief scientist with an $18 million salary Business Insider
Discussion
Sign in to join the discussion
No comments yet — be the first to share your thoughts!