JSOL JMAGビジネスカンパニーは ECCE 2026 (IEEE ENERGY CONVERSION CONGRESS & EXPO)に出展します。
ECCEは、電気およびパワーエレクトロニクス分野の国際会議です。
開催概要
| 主催 | IEEE |
|---|---|
| 日時 | 2026年10月4日（日）～8日（木） |
| 場所 | Vancouver Convention Centre (Vancouver, Canada) |
| ブース | 306 |
| URL | https://www.ieee-ecce.org/2026/ |
Tuesday October 06, 2026 (14:00 - 15:40)
Poster11: NVH, Reliability, Thermal and Materials for Electric Machines
#1242 | Efficient 3D AC Loss Analysis of CORC Cables with Detailed Layering
Yohei Watanabe, Hirokatsu Katagiri, Hiroyuki Sano, Takashi Yamada
To ensure the reliability of Conductor on Round Core cables, accurately evaluating 3D current redistribution and AC losses is important. However, detailed 3D modeling including all cable layers is computationally demanding with conventional approaches. This paper addresses this challenge by adopting the A-φ formulation, enabling efficient large-scale simulations of detailed layered structures. By leveraging this method, models with nearly one million elements are analyzed within a practical timeframe. The results show a trade-off between hysteresis and coupling losses depending on the helical pitch; specifically, a longer pitch promotes current shunting into the stabilizer near tape gaps, increasing coupling loss while mitigating superconducting (SC) layer loss. These findings provide useful insights for SC power applications, demonstrating the effectiveness of the A-φ-based 3D analysis in practical device design.
Tuesday October 06, 2026 (14:00 - 15:40)
Poster13: Reliability and EMI issues in Motor Drives
#1437 | Comparative Study on Motor Model Fidelity for Bearing Current and Common-Mode Voltage Analysis
Hiroyuki Sano, Ryo Endo, Takashi Yamada
This paper proposes a high-fidelity motor modeling approach to analyze bearing currents and common-mode voltage in high-speed wide-bandgap-driven inverters. While rapid switching improves efficiency, it excites parasitic components, causing bearing degradation. To capture these high-frequency phenomena, a distributed parameter model is developed using electrostatic FEA to determine parasitic capacitances for individual wire strands. This model is compared with a simplified version that assumes uniform winding potential. Co-simulation results between electromagnetic FEA and circuits reveal that the proposed model captures critical resonance peaks up to 100 MHz, which are entirely overlooked by the simplified model. The study demonstrates that strand-level modeling is essential for accurately predicting high-frequency energy paths and ensuring the reliability of next-generation motor drives. These findings provide a vital design guideline for mitigating EDM-related failures in SiC and GaN applications.
Wednesday October 07, 2026 (16:05 - 17:45)
C4L-09: Simulation, Surrogate Models, and AI in Motor Design
#1487 | Performance Variation of Machine-Learning-Based Optimization Over Different Electric Machine Design Problems
Shogo Asahino, Kenta Kato, Hiroyuki Sano, Koji Tani, Takashi Yamada
Machine learning (ML)-based optimization offers a significant advantage in electric machine design by enabling model reuse across multiple optimization tasks. However, the robustness of the reused models under different constraint scenarios remains unexplored. This paper investigates the performance of ML-based optimization across three design scenarios for an interior permanent magnet synchronous motor, transitioning from an initial sizing task to detailed multi-physics constrained tasks. A comparative analysis reveals that constraint satisfaction performance is governed by the relationship between optimization objectives and constraints rather than problem scale. When the objective forces the search against the constraint boundary, even small underestimations of constraints by the ML models are exploited, demanding an order of magnitude more training data to achieve local precision. Conversely, when constraints are orthogonal to or aligned with the objective trends, high feasibility is achieved with minimal data. This task-dependent instability demands rigorous post-optimization validation and uncertainty-aware formulations for reliable industrial deployment.
