Integrating FEA into

Existing Process
Flows for Electrical
Machine Design
Presenter: Scott Stanton
Ansoft Corporation

Contributing Author:
Johannes Germishuizen

Siemens AG –
Automation and Drives
A&D LD TD2
 Motivation
Desire to improve existing process flow:
Due to the difficulties in obtaining accurate
analytical solutions
Iterative process needed for transient FEA
 Existing Process
Torque (N-m)

Customer Requirements

Create Initial Design:

Each iteration requires a - Transient FEA
- Voltage Sources
transient solution - Including Field Weakening
- Iterate to obtain δ
 Proposed Process
Torque
Current
Voltage

Create Initial Design.

Customer Requirements Map of solution domain
Scale T, ψd, ψq = f(id, iq,Θ )
N, L using Static Solver

Scale

N, L
 Proposed Process
Verify with Transient
Torque
Current
Voltage

Create Initial Design.

Customer Requirements Map of solution domain
Scale T, ψd, ψq = f(id, iq,Θ )
using Static Solver
N, L

Scale

N, L
User Defined GUI:
Input Parameters
Stator Geometry

Winding Definition
Materials
Rotor Geometry

Solution Range
Maxwell Script: Model Setup

Create Geometry
Assign Materials
Set Boundaries and Excitation
Transform d, q to a, b, c
Setup Torque Calculation
Setup Mesh Operations
 Ansoft Scripting Environment
Ansoft’s products communicate with any program
that utilizes COM (Common Object Module):
Matlab
Excel
Tcl/Tk
Java Script
Visual Basic
Perl
Maxwell Project Setup
Nominal problem utilizes
mesh operation and
Automatic Adaptive Meshing

The Parametric problem reads

in the final adaptive mesh from
Nominal problem, thus no re-
meshing
 Solution Domain
d_min < d < d_max by ∆ dq
q_min < q < q_max by ∆ dq

Where: min, max and ∆ dq are user defined

iA cos θ − sin θ 1 id
iB = cos(θ − 120) − sin(θ − 120) 1 * iq
iC cos(θ + 120) − sin(θ + 120) 1 0
i A = (id cos θ − iq sin θ ) * N
iB = (id cos(θ − 120) − iq sin(θ − 120) )* N
iC = (id cos(θ + 120) − iq sin(θ + 120) )* N
 Solution Domain
-900 < d < 900 : ∆50 1296 variations
-900 < q < 900 : ∆50

Solve

Sequentially Distributed Solve

within Script Option (DSO)

For stage 1 of the project, Θ = 0

Distributed Solve
 DSO – Distributed Solve Option
Host

Nodes …
…
Number of Variations: 1296
Computation Characteristics Sequential Solve: 13 Hr 15 min
Real time 2 Hr 30 min HP Cluster: 8 cpu’s
Speed-up factor 5.3 CPU type: Intel Xeon
Mesh size 6100 triangles 1.87 GHz
2 Gig RAM/CPU
OS: Windows XP
Maxwell Script: Post Processing

Extract Torque Value

Calculate Flux Linkage
Calculate Avg. Tooth Flux Density
Calculate the Normal Component of
the Flux Density in Stator Yoke

Used for Core Loss Calculation

 Post Processing: Torque
One of the advantages of using Maxwell is the
torque is calculated using the virtual work
method:
dW (Θ, i ) d H
TB = |i =const = ( ( B • dH )dV )
dΘ dΘ v 0

As opposed to the analytical method:

3
T = p (ψ m + ( Ld − Lq )id )iq
2
It’s difficult to obtain Ld and Lq analytically
Post Processing: Torque
A,B,C Flux Linkage Calculation
Az|Coil _ a1 dΩ1 − Az|Coil _ a1R e dΩ1
Ψa1 =
dΩ1

ΨA = (Ψa1 + Ψa 2 + Ψa 3 + Ψa 4 ) * SF / ckt
… assuming four coils in solution domain

Where: SF = Symmetry Factor

ckts = number parallel circuits
Maxwell Field Calculator
 D-Q Flux Linkage
dq0 transformation:
ψd cos θ cos(θ − 120) cos(θ + 120) ψA
2
ψ q = * − sin θ − sin(θ − 120) − sin(θ + 120) * ψB
3 1 1 1
ψ0 ψC
2 2 2

User requirement
Maxwell Matrix
 Permanent Magnet Flux
Set d = q = 0 amps and calculate flux linkage of
the Phase A winding: ΨAm

ΨM = ΨA − ΨAm

ΨM ,Ψd ,Ψq , T = f (id , iq )

Output Quantities
 Core Loss
Automatically search
Yoke and Tooth for the
maximum value of B.

5 Bmag dln
Bt1 = ÷5
n =1 dln

→
B • dl1
B y1 =
dl1
 Core Loss

Calculate Core Loss:

1st for Bt_avg_max
2nd for Bymax
Total Loss = {P(Btmax)+ P(Bymax)}*Mass
 Calculate Performance Curves
Solution Domain: Ψd ,Ψq , T = f (id , iq )
{(id1, iq1), (id2, iq2), . . . (idn, iqn)} T=T
T* user required torque
1. Find the minimum stator current:
id|T=T = f(iq) a 1D interpolation to find I1,min
2. With the pair (id, iq)|I1=Imin known, obtain d and q

3. Obtain ud and uq :
ud = R1id − ωψ q
uq = R1iq − ωψ d
 Calculate Performance Curves
4. Calculate the line voltage, U1.
5. Test if U1 exceeds the maximum allowable
voltage, i.e. U1 < Umax (DC bus voltage)
6. If U1 > Umax decrease id until U1 = Umax.
7. Calculate Stator Current and Line Voltage

I1 =
id2 + iq2
U1 =
(
3 ud2 + uq2 ) U1 < U max
2 2

USER DEFINED PROCESS FLOW!

Output of User Defined Process
Flow: Performance Curves

Maxwell solution domain: per turn and per length

User can change N and L to meet inverter requirements
Verification – Maxwell Transient
 Conclusions
Maxwell easily plugs into existing process flow
Full automation
Magnetostatic solver can give reasonable
fingerprint of motor performance
Transient is employed for verification
Proposed process flow greatly increases: Speed,
Accuracy and Capacity over existing process flow.
Appendix: RMxprt Maxwell 2D
Automatic Generation of 2D Design
Appendix: RMxprt Maxwell 2D
Appendix: RMxprt Maxwell 2D
Thank You