JMAG Users Conference
Japan 2026

Electrified powertrains and e-machines present increasingly demanding structural dynamics and NVH challenges, driven by high electromagnetic force harmonics and stringent noise targets. Addressing these issues requires close integration between electromagnetic and structural-dynamic simulation domains, which are traditionally handled by separate tools with limited data compatibility.
To meet this need, we combine JMAG for electromagnetic pre-processing with EXCITE M (AVL), our structural dynamics and NVH simulation tool. EXCITE M’s electromechanical coupling relies on a magnetic field computation based circuit model that represents the machine through multi-dimensional look-up tables of flux linkage, stator tooth forces and torque, parameterized by rotor position and d-q currents, and radial and azimuthal rotor displacement in the case of eccentricity. To connect these domains efficiently, we developed a dedicated interface that generates and transfers the required look-up table data directly from JMAG's magneto-static FE simulations into EXCITE M, eliminating manual data handling and post-processing.
This presentation introduces the coupling methodology, the interface architecture, and demonstrates the resulting workflow on a representative e-machine application, including distinct effects such as dynamic eccentricity, transient effects, and PWM.

Dimensional variations of parts, assembly variance, and other manufacturing tolerances affect motor performance. Quantitatively evaluating these effects during the design stage can identify causes and potential countermeasures before production to help reduce rework during development. This case study configured an analysis model for a motor that accounts for foreseeable variations to aggregate and analyze data under numerous manufacturing tolerance conditions. We also extracted the characteristic modes of variation caused by manufacturing tolerances from the data obtained through analyses and examined whether it is possible to identify the relationship between each mode of variation and the variations in motor performance. This presentation explains our analysis method and findings.

Conventional, parametric design approaches limit the solution space to predefined geometric parameters. Topology optimization expands this space through the free distribution of material within a design domain, enabling more capable designs. Since this requires significantly more designs to be calculated, a larger number of licenses (PSL vs standard) is needed to reach an optimal result within a reasonable optimization timeframe.

In the application scenario "Air Pocket Optimization," the optimal distribution of air pockets within the rotor is investigated with the magnet in a fixed position. The model setup, objective functions, and boundary conditions are presented, along with an evaluation based on a reduced-order model using version 24.2.

Finally, the topology-optimized design is compared with the parametric reference design based on key KPIs. The conclusion and outlook focus on the combined optimization of air pockets and magnet position.

Recent advances in reinforcement learning and imitation learning are accelerating the deployment of robots that operate in close proximity to humans. Toyota has been developing a variety of robotic systems, including the basketball robot CUE and ELEY, a mobile manipulator capable of learning and reproducing human tasks through demonstration. To enable safe and agile operation in human environments, these robots require actuators that simultaneously achieve high backdrivability for compliant physical interaction and high instantaneous torque for human-like dynamic motion.
This presentation introduces our motor development efforts aimed at meeting these challenging requirements. Focusing on Halbach-array outer-rotor motors, we discuss design approaches for balancing competing objectives such as torque output, rotor inertia, cogging torque, and resistance to demagnetization. Practical design examples, prototype evaluation results, and lessons learned through electromagnetic analysis are presented. In addition, future directions for motor technologies supporting next-generation human-centric robots are discussed.

In recent years, the practical adoption of next-generation power devices using SiC and GaN technologies has accelerated the increase of switching frequencies in power converters. This trend is expected to improve converter efficiency and reduce the size.
On the other hand, transformer design for power converters faces issues such as the impact of parasitic capacitance on EMI characteristics and design constraints due to miniaturization. Considering these factors during the design process often requires significant engineering effort.
This presentation introduces a design exploration case study that utilizes optimization techniques incorporating transformer parasitic capacitance, together with coupled electromagnetic, thermal, and structural analyses, to achieve an optimized transformer design.

In electric motor development, with the growing risk of rare earth supply shortages, design risks related to the impact of demagnetization on performance are increasing. Meanwhile, conventional design processes are limited to “verifying the performance impact of switching to alternative magnets,” and optimizing designs using these alternatives requires significant effort.
This presentation will introduce a case study demonstrating how to perform optimization calculations for magnet shapes based on the characteristics of each magnet grade by utilizing JMAG’s optimization and thermal analysis functions.
In addition, we defined how to use optimization functions during the concept and detail design phases as part of our efforts to standardize our motor design process. We will also present a case study on the development of an optimization design process.

［TBA］

N-T characteristics are a critical design benchmark in motor development. However, the design process typically relies on engineers in charge of designing the magnetic circuit to not only run the magnetic field analyses but also organize and post-process the analysis results to obtain the N-T characteristics. To simplify this process, we configured an environment using JMAG scripts that automates N-T characteristic calculations. Our environment takes advantage of JMAG-Express rather than JMAG-Designer as well to enable even those new to JMAG to specify the condition settings and run evaluations. The framework launches the JMAG scripts to automatically calculate the N-T characteristics when JMAG-Express loads the analysis results. One single workflow from running evaluations to calculating the N-T characteristics minimizes the workload and makes calculations previously limited to engineers responsible for the magnetic design accessible to other designers and engineers. This case study provides an overview of the calculation environment, the configuration process, and specific use cases.

High power density is increasingly required for traction motors. To achieve this, it is important to consider not only electromagnetic performance but also thermal and structural characteristics during the design process. However, conventional design approaches require electromagnetic, structural, and thermal analyses to be performed independently with repeated feedback among them, resulting in increased design effort. This presentation introduces a case study on improving design efficiency for automotive traction motors through multi-physics optimization that integrates electromagnetic, thermal, and structural finite element analyses.

IHI strives to capitalize on Model-Based Development (MBD) across multiple technical areas to optimize product designs while considering system-wide tradeoffs. However, Finite Element Analysis (FEA) and other high-fidelity simulations are essential for adequate accuracy in some technological domains. This computational load becomes a bottleneck to running optimizations using one-dimensional models at the conceptual design stage.
This presentation proposes an adaptive optimization for motor design in electrified systems. The process adopts surrogate models trained online through FEA results in JMAG to adaptively switch a portion of the FEA run to surrogate models. These adaptive optimizations not only run faster but also sustain the design solution quality. We will look at some real-world examples and the effectiveness of adaptive optimizations during this presentation.

This presentation introduces an approach to building an Agentic AI system for magnetic simulation, with JMAG and large language models (LLMs) serving as complementary foundations. JMAG provides physics-based modeling and numerical simulation, while LLMs support language understanding, task planning, and engineering reasoning. Reusable skills capture domain knowledge and engineering procedures, providing a structured basis for translating engineering intent into executable tasks. The proposed framework supports integration with workflows across the product lifecycle—from customer requirements analysis and product design to product testing and support for customers’ end-use applications.
This approach aims to streamline simulation operations while broadening engineers’ perspectives on design and development. It supports engineers in exploring alternative design concepts, developing new technical approaches, and assessing their feasibility through simulation.

Axial flux motors offer high design flexibility due to their three-dimensional structure, making the application of optimization analysis challenging. This presentation introduces case studies of electromagnetic and optimization analyses for axial flux motor design. In addition, a material selection study using a newly available feature that treats materials as discrete variables is presented, and the potential of applying optimization analysis to material selection is discussed. Furthermore, examples of optimization studies that consider not only motor performance but also factors such as cost as objective functions are introduced.

Iron loss properties of materials alone cannot precisely predict losses in DC-biased reactors because the DC magnetization distorts the hysteresis loop. This case study uses a play model to estimate the DC-biased B-H characteristics of our MH20D pure iron powder core material as well as run hysteresis loss analyses. These analysis results that closely match measurements prove the effectiveness of our approach for loss prediction in DC-biased reactors. This case study also took advantage of this analysis method to evaluate and optimize reactor designs capitalizing on the unique characteristics of pure iron powder magnetic cores. Our presentation will show some examples of the material selections and reactor designs used for these analyses as well.

In recent years, optimization has become important in induction motor design to improve efficiency and satisfy various design requirements. However, FEA-based optimization requires significant computation time, and the computation time increases rapidly as the number of design variables increases. In addition, nonlinear characteristics caused by magnetic saturation and leakage flux can reduce the prediction accuracy of surrogate models and make optimization results less reliable. This presentation presents an approach to accelerate the optimal design of induction motors by developing a surrogate model trained using historical design assets. It also reports evaluation results demonstrating improved surrogate model accuracy and applicability to optimization through reducing prediction errors caused by nonlinear characteristics.

We are engaged in the research and development of high-temperature superconducting induction motors featuring squirrel-cage rotor winding made of high-temperature superconductors. To date, we have demonstrated outstanding characteristics, including a peak efficiency exceeding 99.5% and rotational stability during rapid acceleration and deceleration. Furthermore, we are advancing multiple application development projects based on these fundamental research findings through industry-academia collaborations; for instance, our high-temperature superconducting submerged pump for liquid hydrogen has reached a practical level of viability.
　　To facilitate the practical implementation of high-temperature superconducting induction motors, precise analysis and design technologies must be established. The characteristics of high-temperature superconductors are defined by a constitutive equation relating electric field and current density; therefore, it is essential to establish an analysis method that accurately incorporates this nonlinear equation.
　　This presentation introduces electromagnetic field analysis results for a practical high-temperature superconducting induction motor, obtained using JMAG®-Designer while accounting for nonlinear electric field vs. current density characteristics. It also compares these analysis results with experimental data and discusses their quantitative accuracy. Finally, based on these findings, the presentation offers an outlook on the potential and future of high-temperature superconducting induction motors.

Stationary-type magnetic refrigeration systems for hydrogen liquefaction are attractive because they operate without moving components. However, the requirement to energize superconducting coils with alternating current results in AC losses that significantly affect the thermal load and overall system efficiency. This presentation reports the evaluation of AC losses in superconducting coils for a magnetic refrigerator using JMAG. The validity of the analysis methodology was verified through comparison with experimental coil results, and a conceptual study of coil models for a future-scale system is also presented.

Magnetic geared motors employ a magnetic gear based on the magnetic field modulation principle, enabling low-speed, high-torque operation without a mechanical gearbox. Their high torque density, reliability, low noise, and oil-free operation make them attractive for mobility applications such as automotive, marine, and railway propulsion systems.

This presentation introduces the development of a 6,000 Nm-class magnetically geared motor using JMAG. Electromagnetic analyses were conducted to evaluate back-EMF, torque characteristics, and losses, including the in-plane eddy current losses in the pole-piece rotor. Through appropriate modeling techniques, a simulation-based design process capable of accurately predicting motor performance prior to prototyping was established.

Experimental results obtained from the prototype showed good agreement with simulation predictions in terms of back-EMF, torque, losses, and temperature rise. The presentation highlights both the analysis methodology and the experimental validation, demonstrating that simulation can provide performance evaluations equivalent to actual machine testing.

The conventional method to analyze the operation of machines that leverage electromagnetic force uses JMAG to obtain the electromagnetic force of parts in specific orientations for import into RecurDyn multi-body dynamic analysis software. However, this approach cannot precisely capture variations in the electromagnetic force as the orientation of the parts changes, lowering the analysis accuracy. Our solution to maintain accuracy configures a coupled analysis that shares the electromagnetic force and part orientation between JMAG and RecurDyn through small time steps. This case study outlines that analysis method and demonstrates how the analysis improves accuracy by comparing the analysis and measured results.

In superconducting coils, accurate evaluation of transient phenomena such as current decay, heat generation, and temperature rise after a quench is important for coil protection and the design of operating conditions. In this study, the quench transient behavior of a superconducting coil was numerically analyzed using JMAG. An analysis model was developed to evaluate the temporal changes in coil current, dissipation of electromagnetic energy, heat generation, and temperature. The validity of the numerical approach was also examined through comparison with experimental results. In addition, as an extended study, a system including a magnetically coupled secondary coil was analyzed to investigate the influence of energy transfer on the transient response during a quench.

As electronic devices continue to operate at higher frequencies, the need to predict inductor frequency characteristics during the design phase is increasing. This presentation introduces an evaluation of broadband impedance characteristics of an inductor in the MHz frequency range using JMAG's full-wave analysis capabilities. The applicability of the proposed approach is discussed with reference to selected measurement results.

Motor requirements for next-generation robots and mobility systems are first identified. A concentrated-winding SPMSM is then optimized for minimum weight under multiple combinations of continuous- and peak-torque requirements, considering electromagnetic, thermal, and demagnetization characteristics. The relationship between the torque requirements and the minimum achievable motor weight, together with the governing constraints, is clarified and summarized as a motor weight characteristic map for machine and motion design.

Recently, demand for Synchronous Reluctance Motors (SynRMs) has been growing to achieve high efficiency and mitigate rare-earth price volatility and supply risks. Although parameter and topology optimization are widely applied to flux barrier shapes governing SynRM performance, the impact of manufacturing variations is often neglected. Conventional optimization considering variations requires repeated Finite Element Analysis (FEA) via Monte Carlo simulations, incurring prohibitive computational costs.
To overcome this challenge, we propose an offline optimization workflow that combines a Gaussian Process Regression surrogate model with Polynomial Chaos Expansion. Using SynRM design as a case study, this presentation reports an optimization that accounts for manufacturing variations while reducing computation time to less than 1/10 of the conventional Monte Carlo method.