JMAG Newsletter July, 2015Product Report

[Back]
Introduction of JMAGDesigner Ver.14.1

JMAGDesigner Ver.14.1 (hereafter Ver.14.1) was released on June, 2015.
Operability of the Geometry Editor and post processing function has been improved. The new added functions will reduce step, straightforward operations will allow efficient analyses. Analyses incorporating specific phenomena such as iron loss accounting for the effects of residual strain, and global optimization calculations using the optimization engine of MATLAB can both be performed. This document will be introducing the new features of Ver.14.1

Introduction
JMAGDesigner Ver.14.1 (hereafter Ver.14.1) was released on June, 2015.
The operability of the Geometry Editor and the postprocessing function has improved in performance with Ver.14.1. Iron loss analysis accounting for production degradation, optimization function linked with MATLAB, as well as new modeling technology and evaluation analysis function have been developed for the purpose of performing detailed analyses and to avoid designs missing vital factors.
Improvements in the user interface of the Geometry Editor and the postprocessing function have enhanced in efficiency and reduced analysis times. The latest version has reduces the number of necessary steps and improved the operability by incorporating a new mechanism.
Using the improved Geometry Editor, 3D models that have simulated specific details in the geometry can be easily created, in addition to the creation of 2D models. The automatic healing and mesh generation functions have been enhanced for the evaluation of these models. With these functions, analysis of magnet eddy current loss in PM motors and analysis of stray loss that occurs in the case of transformers (both require evaluations using 3D models) can produce stable results.
As for the material modeling aspect, the function to account for deterioration in steel sheets due to punching can now analyze increases in iron loss due to deterioration and decrease in torque performance.
To search through a larger and complicated design space, a global optimization engine that supports the multimodality issue is necessary. The MATLAB global search optimization engine can now be used directly. In addition, JMAG has also developed an optimization engine using genetic algorithm, and practicality of the optimization function has improved.

Improved Operation
The toolbar buttons have been organized and rearranged for easier selection. Also, a button has been added to reduce the number of clicks. As a result, creating geometries and postprocessing processes are made easier and faster. This allows reduction in time necessary in the entire analysis operation.
Improved Operability of the Geometry Editor
The Geometry Editor has been improved for easier operation with fewer steps. Despite the increase in toolbar buttons with the addition of functions, the interface has been sorted for efficient operability. We will be introducing two major improvements in Ver. 14.1.
1. Only displays required functions
The new Geometry Editor has a mechanism that will switch the displayed toolbar buttons depending on the edit mode, such as edit sketch and edit part, allowing easier access to necessary tools. Unorganized buttons have been recategorized by theme, and the user interface is friendlier than it was before (Fig.1).
Fig.1 Toolbar button display when the Edit Part button is turned on
Only the toolbar buttons that can be used in the edit part mode is displayed in the frame.
The menu displayed when rightclicking on the entity has also been enhanced. In previous versions, rightclicking on a region displayed tools that seemed unnecessary in region edit, such as point and line creation. Ver.14.1 has been improved to only display functions that can be used for regions, such as region mesh generation and the region copy function in the menu (Fig.2). Necessary functions are now easier to be selected and operated through this improvement.
Fig. 2 Rightclick menu for the region
Only displays mesh generation and region copy tools used for regions
2. Mechanism to reduce the number of clicks
A single region can be created with a small number of clicks, but the same would not apply for 100 regions. Just reducing one click for a region makes all the difference. The Apply button that was incorporated for converting sketches and creating fillets in Ver.13.1 has been added also for creating regions and generating manual mesh (Fig.3). This has realized continuous processes with minimal clicks.
Fig. 3 Apply button added for region mesh generation
Mesh generation is continuously possible for 5 regions using the Apply button.
Element size settings are succeeded and the number of clicks for mesh generation has been reduced.
Improved Constraint Function
Geometry parametric analysis can be run by setting the appropriate constraint functions. In addition to the function to display excess constraints with the color orange in Ver.14.0, a function to display perfect constraints in black has been added in Ver.14.1 to get a better grasp of the constraint state (Fig.4).
Fig. 4 Differences in complete constraints and constraint display
(1),(2) highlighted in black have their end point and direction constrained, making it perfectly constrained.
(3) is only direction so it is a regular constraint shown in blue.
Perfect constraint is the state that all related geometrical constraints and dimension constraints are defined. For example, the line of (1) displayed in black within Fig.4 has fixed constraints on the end point, determining the position, direction, and distance. The two lines of (2) are determined in position and direction. For the line of (3), only the direction is determined, so it is in the regular blue constraint state. Geometry parametric analysis can be run without collapsing the geometry by putting a perfect constraint on the lines and arcs that configure the geometry, as well as determining the shape.
Simultaneous Output of Analysis Results for All Cases
A new function has been added to collect values of all cases for torque and loss used in the graph after calculating multiple cases in parametric analysis, then outputting them all to one csv file.
For example, let's look at results where torque variations of all cases have been output after running a parametric calculation with changed current phase (Fig.5). Selecting [Torque] in the output information prior to calculation and specifying the folder path allows the output of all torque variations per case to one csv file. Importing all result files in JMAGDesigner and exporting the created table of torque variations can be omitted, allowing reductions in time for postprocessing.
This function can be used in distributed processing mentioned later. After distributed processing, specifying the path of the shared file system that can be accessed by the calculation node allows collection of important information to csv files.
Fig. 5 Output csv data
Actual example of output data. The torque variations of all calculated cases are organized in case order.
Data of all cases can be confirmed after calculation is completed.
General Improvements in the Mesher
Mesh generated for complicated 3D models with a grasp on geometry details can be obtained with enhancements and improvements in the algorithm of mesh generation and the healing function.
With improvements in the algorithm, the success rate for mesh generation has improved for models with parts of various scales and models including numerous parts. Mesh generation is more stable than ever, with the removal of issues in the CAD model by adding a healing function that can control geometry modifications. Here, we will be introducing enhancements in the healing function.
Enhanced Healing Function for Mesh Generation
Mesh can easily be generated for geometries including gaps between parts, interference, small faces and small edges by adjusting the tolerance of geometry modification.
Geometry models that have been imported from the Geometry Editor or other CAD may include displacements of parts or small geometries depending on the dimensional errors of the model or compatibility with CAD (Fig.6). The modification range of geometries can be controlled with the new added tolerance settings and modifications of geometries can be adjusted by the user (FIg.7), As a result, mesh can be created and calculation can be run without having to return to geometry modifications in the CAD (Fig.8).
Fig.6 Model intervened by the coil and core
Model where there is an interference of 0.001mm in the highlighted area out of the areas where the coil and core are in contact.

Fig. 7 Specification of tolerance to run healing
Specifies length tolerance of 0.002mm that is longer than the interference of 0.001m extracted in Fig.6, and runs automatic healing during mesh generation.

Fig. 8 Successful mesh generation
Interference less than tolerance that is automatically modified has been highlighted, and the modified area can be confirmed after mesh generation.


Enhancements in Modeling Technology
In terms of modeling technology, we are working hard towards development in incorporating factors required in detailed loss analyses. Models that account for the effects of material properties that affect loss, such as production degradation in the magnetic steel sheet, have been added.
Other than highprecision loss analyses, new functions have been added for detailed modeling of temperature rise due to loss, and phenomena such as vibrations and noise caused by electromagnetic force. The multipurpose file import / export tool has been improved to allow incorporation of vibration analyses that perceive the mover and the stator as the vibratory source.
Material Modeling for Production Degradation
We improved the analysis functions for evaluating iron loss increases and torque performance decreases due to residual strain occurring during electromagnetic steel sheet punching. The degree of degradation facing the innerside from the cut plane can be identified, and residual strain effects can be grasped by using a function to automatically generate layer mesh expressing product degradation regions, and models which automatically assign degradation properties of materials corresponding to the distance from the cross section for each layer.
2 layers of 1mm layer mesh are generated from the stator core edge (Fig. 9). By adding the information of the degradation degree for the distance to every magnetic properties and iron loss properties of the material database used in the stator core, the material properties accounting for degradation are automatically defined in the degraded regions.
In this manner, by accounting for iron loss increases accumulating in the steel sheet edge due to punching, even more accurate loss amounts can be obtained (Fig. 10).
Fig. 9 Production degradation region layer mesh
Append residual strain condition to steel sheet edges which have been punched and automatically generate 2 layers of mesh with a specified width of 1mm.

Fig. 10 Hysteresis loss distribution accounting for production degradation
Hysteresis loss increases in the degraded region 2mm from the punched edge, and concentrates especially at the edges of teeth where magnetic flux density is large.

Added Functions to the Material Database
Kobe Steel soft magnetic composites, Hitachi Metals/Arnold neodymium sintered magnets, ferrite, and Arnold SmCo and Alnico magnets have been added (Table 1, 2). Now 226 types of core materials and 534 types of permanent magnet materials are available in the material database.
There are also an increased number of material data import formats. Text file material data exported from PowerCore® Explorer, a material data management tool developed by ThyssenKrupp AG, can be imported into a custom material. BH curves and iron loss curves can be imported (Fig. 11).
Table. 1
Number of core materials added to material data and total number of materials
Material type 
Soft magnetic composite 
Kobe Steel 
4 
Total number of materials 
12 

Table. 2
Number of permanent magnets added to material data and total number of materials
Material type 
Neodymium Sintered magnets 
Ferrite 
smco 
alnico 
Hitachi 
6 
2 
 
 
Arnold 
74 
7 
20 
18 
Total number of materials 
221 
119 
25 
38 

(a) PowerCore® Explorer iron loss data

(b) JMAGDesigner database

Fig. 11 Example of imported iron loss curves
Import iron loss frequency characteristics data from PowerCore® Explorer(a) into the JMAG database (b) via a text file.

Improved Multipurpose File Import/Export Tool
We improved the function for mapping electromagnetic force distribution to structural analysis models. This improvement allows electromagnetic force to be automatically delivered to both movers and stators, and a vibrational analysis can be performed.
In many cases a periodic model is used in a magnetic field analysis. However, a full model is used instead of partial model in a structural analysis, since parts with no symmetry such as housing need to be modeled and they cannot be handled as a partial model. As for periodic conditions and antiperiodic conditions, mapping the analysis results to full model in a structural analysis has already been supported, but mapping for axial reverse rotational periodic conditions is also now supported.
Adding these two functions allows vibrational analysis of claw pole alternators to be performed easier than before (Fig.12).
Fig. 12 Vibrational analysis of a claw pole alternator
Deformation and displacement distribution when performing a vibrational analysis of an alternator based on the electromagnetic force working in the rotor area of a magnetic field axial reverse rotational periodic model.
Enhanced Abaqus Coupling Interface
The direct interface for Abaqus and JMAGDesigner has been improved. Ver.14.1 now allows large deformation analysis to be performed with electromagnetic force.
The time intervals for Abaqus' explicit method procedure (Dynamic Explicit), which is required in a large deformation analysis, are too short in the transient response magnetic field analysis, and when these intervals are used in a magnetic field analysis, the number of calculations increases enormously. The subcycling coupling is now supported in this latest version, and a twoway coupled analysis can be performed without requiring the time intervals to match in Abaqus and JMAGDesigner. Calculations are more stable, and electromagnetic force and displacement can be passed with the required timing, enabling a coupled analysis with large deformations, such as an electromagnetic forming analysis for thin plate (Fig. 13). Thin plate deforming in the direction of the mold due to electromagnetic force (vector in the figure) can now be captured.
Fig. 13 Electromagnetic forming analysis of a thin plate
Example of performing a twoway coupled analysis with JMAG's transient response magnetic field analysis and Abaqus' Dynamic Explicit, and then analyzing the deformations of a plate.
Thin plate deforming in the direction of the mold due to electromagnetic force shown in the vector can now be verified.

Enhanced Evaluation Functions
As a function to analyze and optimize results, new evaluation functions have been added, and the optimization engine has been improved.
The frequency filter function added to the evaluation method of the transient analysis results visualizes specific frequency components or superimposed distribution amounts, which are difficult to grasp in the time axis. Evaluations connecting the positional relationship between geometries and the associated frequencies with distribution amounts contribute to improvement of design plans.
The MATLAB optimization engine can now be used directly, and the operability of optimization function has been improved. Distributed processing functions have also improved so calculations for several thousand cases associated with optimization can be efficiently performed. Direct job implementation for generalpurpose batch systems such as LSF and PBS is now supported, and functions have been added for easier optimization using the existing environments.
Frequency Filter Function
The contours and flux lines of magnetic flux density where harmonic analysis has been carried out can be filtered for only specific frequency components and displayed in the time axis again. Including the positional relationship between the slot and magnet, the harmonic components of magnetic flux that may cause the magnet eddy current can be evaluated.
Let's look at the example of magnet eddy current generated in an IPM motor. Large eddy currents are generated locally in the magnet (Fig. 14). By running a Fourier transformation directly from the magnetic flux density graph obtained from the point with large eddy current, it's possible to confirm which frequency component is causing the local eddy currents (Fig. 15). By extracting only large components as a time variation component of magnetic flux density and looking at the magnetic flux lines on a time axis, the effects that the positional relationship between the slot and the magnet have on the eddy current of the magnet can be verified (Fig. 16). By accounting for the magnetic flux delay occurring in the stator and displaying it including before and after the frequency orders focused on, magnetic flux density wrapping around the teeth can be simultaneously verified.
Fig. 14 Magnet joule loss distributoin
Joule loss distribution occurring in the magnet. We learn that not only in the ends of the magnet, but also a comparatively large amount of losses are also occurring in the center section.

Fig. 15 Magnetic flux harmonic analysis
Fig. 14 Harmonic analysis results of magnetic flux density in the center of the magnet with large joule losses.
We can see that the component increases significantly at 0Hz due to magnet magnetic flux and increases at 1800Hz (12th degree) again.

Fig. 16 Slot harmonic components for magnetic flux lines
After the 12th order component for the magnetic flux density obtained from the harmonic analysis in Fig 15 is added to the nearby order (11th and 13th) components, the magnetic flux lines are displayed with the time history.
Magnetic flux interlinking areas with increasing joule loss shown in Fig. 14 could be identified.

Added Genetic Algorithm Function
MATLAB optimization tool for global search method can be set in the same way as the existing optimization functions. Genetic algorithms as well as original optimization engines created in MATLAB can be used. Additionally, an optimization engine which used original JMAG genetic algorithms was developed.
Many of the problems for optimization are multimodal. By using the newly added genetic algorithm or MATLAB global search method, optimization can be applied to a wide variety of actual problems. This new optimization engine can be directly used from JMAGDesigner like before, and optimization calculations can be performed simply by setting the objective functions and selecting the genetic algorithm.
Directly accessing the engine through JMAGDesigner enables you to handle a case such as optimizing geometry accounting for the tradeoff between magnetic circuit design and structural design. By defining the constraint condition in addition to the objective function, the objective function can be maximized while keeping the constraints. (Fig. 17). By using the analysis group function, torque can be maximized by setting the maximum value of the stress applied to the rotor as a constraint condition.


Fig. 17 Objective function settings
Define torque maximization as an objective function, and set the Mises stress, acting as limit for strength design, as a constraint condition.

Fig. 18 Example of maximizing torque
At approximately 350 cases, average torque of more than 350Nm can be obtained. This graph shows that various design plans were searched for up to 100 cases from the torque variations, and as the number of cases increases, the design plans are narrowed down to optimum one.

Distributed Calculation using GeneralPurpose Batch Systems
For the HPC server environment to be effectively used, the method of execution was changed so the job could be directly added with LSF, PBS or other generalpurpose batch systems. This saves the trouble of creating specialized scripts; distributed processing using a general purpose batch system can be performed in the same way as a regular local machine.
Obtain the required execution scripts and SSH login information from the server administrator and register the information (Fig. 19)
Fig. 19 Example of adding remote machine information
When the remote login information and execution scripts are registered in advance in [Tools] > [Settings] > [SSH], it is possible to select LSF_server registered in the server section of the batch execution.

Conclusion
We hope you like the new version. We introduced some of the new features available in Ver. 14.1.
To see all of the features, please visit our homepage to view videos of each function (*1). Also, there are sample data for each function (*2).
We hope that all of the latest features in JMAG are useful for you.
(Mari Nakamura)
*1 JMAG Function Videos:
http://www.jmaginternational.com/products/jmagdesigner/index.html#video
*2 Sampledata:
http://www.jmaginternational.com/products/jmagdesigner/index_v141.html
Please access sample data link on Introduction page of JMAGDesigner Ver.14.1. User authentication is required to access there.

Contents
1. Solutions  6th Seminar on Advanced Computational Electromagnetics
Toward Putting HighPrecision Loss Analysis Into Practice Followup 
2. Product Report  Introduction of JMAGDesigner Ver.14.1 
3. Fully Mastering JMAG  Common Questions for JMAG 
4. Event Information



