37
© ika 2018 · All rights reserved 15.11.2018 Slide No. 1 · 18jhe0005.pptx Friedrichshafen, 15. November 2018 Jonas Hemsen M.Sc., Dipl.-Ing. Daniel Kieninger, Prof. Dr.-Ing. Lutz Eckstein Holistic Design of Electric Drive Modules for Vehicles FEMAG Anwendertreffen 2018 Institute for Automotive Engineering RWTH Aachen University source: GM

FEMAG Anwendertreffen2018

  • Upload
    others

  • View
    1

  • Download
    0

Embed Size (px)

Citation preview

Page 1: FEMAG Anwendertreffen2018

© ika 2018 · All rights reserved15.11.2018Slide No. 1· 18jhe0005.pptx

Friedrichshafen, 15. November 2018

Jonas Hemsen M.Sc., Dipl.-Ing. Daniel Kieninger, Prof. Dr.-Ing. Lutz Eckstein

Holistic Design of Electric Drive Modules for Vehicles

FEMAG Anwendertreffen 2018

Institute for Automotive Engineering – RWTH Aachen University

source: GM

Page 2: FEMAG Anwendertreffen2018

© ika 2018 · All rights reserved15.11.2018Slide No. 2· 18jhe0005.pptx

Agenda

Holistic Design of Electric Drive Modules

MATLAB & FEMAG

Calculations:

Parameter Identification

Losses

Cogging torque

Torque ripple

Constraints:

Mechanical load, Back EMF, Short circuit current

Demagnetization of PMs

Thermal behaviour

Outlook

FEMAG as a „Newbie“

Page 3: FEMAG Anwendertreffen2018

© ika 2018 · All rights reserved15.11.2018Slide No. 3· 18jhe0005.pptx

Holistic Design of Electric Drive Modules

Module characteristics result from multi-physical interaction of all drivetrain components

Currently state of the art:

“Optimization” of a single component for existing drive systems.

Design of drive modules with simplified component models E.g.

Constant losses over complete operation range

Ideal torque production of EM (no ripples)

Resulting problems:

Optimization of single component leads to local optimum → Goal: Global optimum

Sequential and iterative component optimization takes long time and is expensive

“The best electric machine, the best inverter and the best

transmission do not necessarily combine to the best drive

module.”

Page 4: FEMAG Anwendertreffen2018

© ika 2018 · All rights reserved15.11.2018Slide No. 4· 18jhe0005.pptx

Holistic Design of Electric Drive Modules

Solution: Optimize and evaluate drive module on system level with detailed component models

Propulsion System Design

Design Conditions

Driving requirements and boundaries

Calculation of component requirements

from driving requirements

Component Level

Electric Machine Design

Inverter Design

Transmission Design

Assembly of components and

identification of best set

1

2

3

5

6

Request of components from component

level blocks

Component requirements

Laid out and calculated

components

Layouting of components which

fulfil requirements4

source: ProFEMAG

Page 5: FEMAG Anwendertreffen2018

© ika 2018 · All rights reserved15.11.2018Slide No. 5· 18jhe0005.pptx

Holistic Design of Electric Drive Modules

Example for identified interactions between components

Electric Machine

• High Speed electric machine

• High power densities

• Power mainly by speed (lower torque)

Transmission

• High reductions ratios

• High tooth meshing

• Big helix angle

Inverter

• High switching frequencies

• Small Capacitor

Compact High losses High switching losses

On system level: Benefit or drawback?

Cost-effective CompactLess compact

Page 6: FEMAG Anwendertreffen2018

© ika 2018 · All rights reserved15.11.2018Slide No. 6· 18jhe0005.pptx

Agenda

Holistic Design of Electric Drive Modules

MATLAB & FEMAG

Calculations:

Parameter Identification

Losses

Cogging torque

Torque ripple

Calculated Contraints:

Mechanical load, Back EMF, Short circuit current

Demagnetization of PMs

Thermal behaviour

Outlook

Conclusion: FEMAG as a „Newbie“

Page 7: FEMAG Anwendertreffen2018

© ika 2018 · All rights reserved15.11.2018Slide No. 7· 18jhe0005.pptx

Complete calculation process:

These operations are explained in more detail on the following slides

1.

User Input

•Matlab GUI

•All Machine Data

2.

Constructionof FEMAG Script

•Combine snippetsto full script

•Manipulatingvariables accordingto GUI Values

•Various scripts fordifferent calculations

3.

Executionof FEMAG Script

•System Command to startFEMAG

•Scriptname ispassed as option

4.

Wait forFEMAG toFinish

5.

Read out raw data

•Either fromFEMAG .BCH fileor created .csv file

•Put data toMATLAB format

6.

Post-processingof data

•e.g. calculation oftorque-speed curve

MATLAB & FEMAG

Page 8: FEMAG Anwendertreffen2018

© ika 2018 · All rights reserved15.11.2018Slide No. 8· 18jhe0005.pptx

MATLAB & FEMAG – 1. User Input

All coding is done in

MATLAB

For the reason of reusability

and structure, a GUI is built

up

manual input or filling of

fields by script

Page 9: FEMAG Anwendertreffen2018

© ika 2018 · All rights reserved15.11.2018Slide No. 9· 18jhe0005.pptx

MATLAB & FEMAG – 2.Construction of FEMAG Script

Result is a runnable .fsl file which is designed according to the GUI inputs including:

Machine definition

Material assignments

FEMAG preprocessing functions

FEMAG calculation method („pm_sym_fast“, „ld_ld_indent“, etc.)

FEMAG postprocessing functions

And much

more….

foo.fsl

Rotor Definition

Winding Head

Definition

Magnet Data

.fsl snippets Complete FSL file Replace variables

in file

Page 10: FEMAG Anwendertreffen2018

© ika 2018 · All rights reserved15.11.2018Slide No. 10· 18jhe0005.pptx

MATLAB & FEMAG – 3. Execution of FEMAG Script

How can we start FEMAG from MATLAB?

Solution: Command line prompt in MATLAB

Script “foo.fsl” as parameter

Same as

Page 11: FEMAG Anwendertreffen2018

© ika 2018 · All rights reserved15.11.2018Slide No. 11· 18jhe0005.pptx

How can we detect, when the calculation is done?

Solution: Periodically checking the „journal“ file (.jrn) for its stop time.

E.g. checking every 1 second for new stop time in a loop

Also FEMAG errors can be detected by checking for the string „ERROR“

Suggestion for the future: Having a mechanism for feedback from FEMAG when calculation finishes

MATLAB & FEMAG – 4. Wait for FEMAG to Finish

Page 12: FEMAG Anwendertreffen2018

© ika 2018 · All rights reserved15.11.2018Slide No. 12· 18jhe0005.pptx

MATLAB & FEMAG – 5. Read out raw data

Two possible methods are

used supplementary:

1. Read-in text based

values from .BCH file

(simplified for reasons of understandability)

Page 13: FEMAG Anwendertreffen2018

© ika 2018 · All rights reserved15.11.2018Slide No. 13· 18jhe0005.pptx

MATLAB & FEMAG – 5. Read out raw data

Two possible methods are used

supplementary:

2. Read in numeric values from

constructed .csv file

(simplified for reasons of understandability)

Page 14: FEMAG Anwendertreffen2018

© ika 2018 · All rights reserved15.11.2018Slide No. 14· 18jhe0005.pptx

MATLAB & FEMAG – 6. Post-processing of data

Following Machine Characteristics can be calculated post-processing of FEM

results

Cogging torque and torque ripple (without skew)

MTPA control parameters (current, voltage, current angle)

Torque-speed characteristic using MTPA

Efficiency diagrams at different temperatures

Losses and Efficiency

Approximate Mass, Volume and Costs

Page 15: FEMAG Anwendertreffen2018

© ika 2018 · All rights reserved15.11.2018Slide No. 15· 18jhe0005.pptx

Agenda

Holistic Design of Electric Drive Modules

MATLAB & FEMAG

Calculations:

Parameter Identification

Losses

Cogging torque

Torque ripple

Calculated Contraints:

Mechanical load, Back EMF, Short circuit current

Demagnetization of PMs

Thermal behaviour

Outlook

Conclusion: FEMAG as a „Newbie“

Page 16: FEMAG Anwendertreffen2018

© ika 2018 · All rights reserved15.11.2018Slide No. 16· 18jhe0005.pptx

Calculations: „Parameter Identification“

FEMAG Method: „ld_lq_fast“

Superior over „psd_psq_fast“ due to definition of I and beta instead

of I_d and I_q

5 beta steps, 5 current steps

Interpolation in-between for smoother results

Good compromise for still accurate enough results and reduced

computation time

Only two rotor positions:

1. Magnet in axis with slot

2. Magnet in axis with tooth3,75 °

Iq

Id

Invalid: 𝐼 > 𝐼𝑚𝑎𝑥

Valid: 𝐼 ≤ 𝐼𝑚𝑎𝑥

𝐼 = 𝐼𝑑2 + 𝐼𝑞

2

𝐼𝑚𝑎𝑥

Page 17: FEMAG Anwendertreffen2018

© ika 2018 · All rights reserved15.11.2018Slide No. 17· 18jhe0005.pptx

Calculations: „Parameter Identification“

1. Number of current and β steps:

High inaccuray when < 5 steps

Also influences almost linearly calculation time

2. Rotor position steps

Acceptable inaccuracy between 15 and 25 steps

Influence of calculation settings

Page 18: FEMAG Anwendertreffen2018

© ika 2018 · All rights reserved15.11.2018Slide No. 18· 18jhe0005.pptx

Calculations: „Parameter Identification“

3. Relative node distance

No significant influence both on calculation time

and accuracy

4. Rotor Move range

No influence

Derived „rule of thumb“

5 current-steps, 5 β-steps

Rotor position steps: 2

Rotor move range: 180°/slots

Relative node distance

Rotor: 4

Stator: 3

Influence of calculation settings

Page 19: FEMAG Anwendertreffen2018

© ika 2018 · All rights reserved15.11.2018Slide No. 19· 18jhe0005.pptx

Calculations: „Losses“

Copper losses: 𝑃𝑣_𝑐𝑢 = 𝐼2 ∗ 𝑅𝑝ℎ𝑎𝑠𝑒

Iron losses: more complex

FEMAG Method: One „pm_sym_fast“ per I-β

combination

25 “pm_sym_fast” calculations for 5x5 I-β

grid

Interpolation in-between for smoother

results

In each “pm_sym_fast”, losses are calculated

for multiple frequencies by FEMAG internal

flux density values.

post_models("calc_losses","FeLoss")

No additional FEM necessary, done by

post-processing

Easier than manual calculation of losses in

each element and summing up.

(simplified for reasons of understandability)

Page 20: FEMAG Anwendertreffen2018

© ika 2018 · All rights reserved15.11.2018Slide No. 20· 18jhe0005.pptx

Calculations: „Losses“

Used Model:

Jordan, but equally possible with all other models

Alternative:

Accessing each and every mesh element by “get_elem_data(identifier, key)

Page 21: FEMAG Anwendertreffen2018

© ika 2018 · All rights reserved15.11.2018Slide No. 21· 18jhe0005.pptx

Calculations: „Losses“

Results: 4-dimensional loss data for each subregion of

the machine

P_fe(I, β, f, region)

Still some optimization potential since some

combination to not occur in real EM operation (e.g.

very high currents at high field weakening (very

small β)

Loss data must be transferred to operation area of

EM (e.g by means of MTPA control scheme)

Page 22: FEMAG Anwendertreffen2018

© ika 2018 · All rights reserved15.11.2018Slide No. 22· 18jhe0005.pptx

Calculations: „Losses“

MTPA (Maximum torque per ampere) / MTPV (Maximum

torque per volt) control

Find minimum current for given torque -> minimizes

copper losses

Consider Current and voltage boundaries

MTPA/MTPV

control

𝐿𝑑(𝐼, β)

𝐿𝑞(𝐼, β)

ψ𝑝𝑚(𝐼, β)

𝑇𝑟𝑒𝑞

source: Schröder, D.

𝐼𝑚𝑎𝑥 ≤ 𝐼𝑑2 + 𝐼𝑞

2 𝑈𝑚𝑎𝑥 ≤ 𝑈𝑑2+ 𝑈𝑞

2

𝐼𝑑

𝐼𝑞

𝑈𝑚𝑎𝑥

𝐼𝑚𝑎𝑥

Page 23: FEMAG Anwendertreffen2018

© ika 2018 · All rights reserved15.11.2018Slide No. 23· 18jhe0005.pptx

Calculations: „Losses“

Results of MTPA/MTPV control:

Transformation of loss data to torque-speed plane

„Raw“ losses from loss calculation Losses mapped into Torque-speed plane

Page 24: FEMAG Anwendertreffen2018

© ika 2018 · All rights reserved15.11.2018Slide No. 24· 18jhe0005.pptx

Calculations: „Cogging Torque“

FEMAG Method: „cogg_calc“

Move steps: 20

Angle range: 360°/slots

Mechanical angle that covers a complete period

Other settings like in parameter identification

Page 25: FEMAG Anwendertreffen2018

© ika 2018 · All rights reserved15.11.2018Slide No. 25· 18jhe0005.pptx

Calculations: „Torque ripple“

FEMAG Method: „torq_calc “

Move steps: 20

Angle range: 2 * 360°/ slots

Mechanical angle that covers two periods

Other settings like in parameter identification

Page 26: FEMAG Anwendertreffen2018

© ika 2018 · All rights reserved15.11.2018Slide No. 26· 18jhe0005.pptx

Agenda

Holistic Design of Electric Drive Modules

MATLAB & FEMAG

Calculations:

Parameter Identification

Losses

Cogging torque

Torque ripple

Calculated Contraints:

Mechanical load, Back EMF, Short circuit current

Demagnetization of PMs

Thermal behaviour

Outlook

Conclusion: FEMAG as a „Newbie“

Page 27: FEMAG Anwendertreffen2018

© ika 2018 · All rights reserved15.11.2018Slide No. 27· 18jhe0005.pptx

Constraints: „Mechanical Load“, „Back EMF“ and „SC current“

No extra FEMAG calculation neccessary

Mechanical Load: Simple model

Limited by circumferential speed (𝑣𝑚𝑎𝑥 ≤ 250 𝑚/𝑠1)

Back EMF at cold magnet temperature

𝐸𝑀𝐹 = 𝑗 ω𝑒𝑙 ∗ ψ𝑃𝑀

𝐸𝑀𝐹𝑚𝑎𝑥 ≤ 550 𝑉

Short circuit current (only in d-direction)2

𝐼𝑑 = 𝑗ψ𝑃𝑀

𝐿𝑑

1: According to Wu, T; 2: According to Finken, T.

Page 28: FEMAG Anwendertreffen2018

© ika 2018 · All rights reserved15.11.2018Slide No. 28· 18jhe0005.pptx

Constraints: Demagnetization of Permanent Magnets1

00

De

g.C

15

0 D

eg.C

25 Rot. Pos. StepsMinimum Rot. Pos.

Steps (13)

% 𝑎𝑟𝑒𝑎 𝐻 < 𝐻𝑚𝑖𝑛 = 29.8%

FEMAG Method: “pm_sym_fast”

Worst case for demagnetization

Max. magnet temperature (here 150 °C)

Current completely in negative d-direction

Calculation time is lower with less rotor position

steps (93 s vs. 214 s)

Rotor position steps can be minimized without

(great) loss of accuracy

Here: Demagnetizing to a high extend

Current must be lowered or…

Magnet size increased or…

Stronger magnets or…

Lower temperature limit

𝐻𝑚𝑖𝑛 = −842 Τ𝑘𝐴 𝑚 ; 𝐼𝑑 = −250𝐴 𝑟𝑚𝑠; 𝑇𝑚𝑎𝑔 = 150°C

% 𝑎𝑟𝑒𝑎 𝐻 < 𝐻𝑚𝑖𝑛 = 32.0%

% 𝑎𝑟𝑒𝑎 𝐻 < 𝐻𝑚𝑖𝑛 = 71.5% % 𝑎𝑟𝑒𝑎 𝐻 < 𝐻𝑚𝑖𝑛 = 71.7%

Page 29: FEMAG Anwendertreffen2018

© ika 2018 · All rights reserved15.11.2018Slide No. 29· 18jhe0005.pptx

Constraints: „Thermal Behaviour“

Co

mp

uta

tio

ntim

e

Exactness High

High

Low

Thermal

FEA

𝑆 < 18 𝐴/𝑚𝑚2

𝐴 < 210 𝐴/𝑚𝑚

Thermal

Network

Possibilities:

Analytical constraints

Current density and -coverage

Thermal FEA

Thermal equivalent network

Best compromise between exactness and

computation time: Thermal equivalent Network

𝐴 =𝑁 ∗ 𝑧𝑛 ∗ 𝐼

𝜋 ∗ 𝐷𝑟𝑜𝑡

𝑆 =𝑧𝑛 ∗ 𝐼

𝐴𝑛 ∗ 𝜑𝑛

𝐴 Current coverage

𝑆 Current density in Slot area

𝜑𝑛 Fill factor of slots

𝑁 Number of slots

𝐼 Current

𝑧𝑛 Turns per slot

𝐷𝑟𝑜𝑡 Rotor diameter

Page 30: FEMAG Anwendertreffen2018

© ika 2018 · All rights reserved15.11.2018Slide No. 30· 18jhe0005.pptx

Constraints: „Thermal Behaviour“

Thermal Equivalent Network – Principle:

Also called „Lumped Parameter Transient Model“

Used to esimate temperature development in machine

parts upon load (e.g. cycle)

Each machine part is considered as „chunk“ with

homogeneous temperature and material.

Problem: different machine topologies need different

networks

TcoolPs-yo Pte Pacwi

PenwiPpm

Pr-coPbe

Cfr Cs-yo Cte Cacwi

CenwiCpm

Cr-coCbe

Rco-fr Rfr-s-yo Rs-yo-te Rte-acwi

Racwi-enwiRte-pm

Rpm-r-co

Rr-co-fr

Rr-co-beRbe-fr Renwi-r-co

Renwi-fr

Coolant Frame Stator Yoke Teeth Active Winding

End WindingMagnets

Bearing Rotor Core

TcoolPs-yo Pte Pacwi

PenwiPpm

Pr-coPbe

Cfr Cs-yo Cte Cacwi

CenwiCpm

Cr-coCbe

Rco-fr Rfr-s-yo Rs-yo-te Rte-acwi

Racwi-enwiRte-pm

Rpm-r-co

Rr-co-fr

Rr-co-beRbe-fr Renwi-r-co

Renwi-fr

Coolant Frame Stator Yoke Teeth Active Winding

End WindingMagnets

Bearing Rotor Core

Thermal quantity Equivalent electric quantity

Thermal resistance 𝑅𝑡ℎ Τ𝐾 𝑊 Electrical resistance 𝑅𝑒𝑙 𝛺

Temperature difference ∆𝑇 𝐾 Electric potential 𝜑 𝑉

Temperature 𝑇 ℃ Voltage 𝑈 𝑉

Heat flow ሶ𝑄 𝑊 Current 𝐼 𝐴

Thermal capacity 𝐶𝑡ℎ Τ𝐽 𝐾 Electric capacity 𝐶𝑒𝑙 𝐹

Page 31: FEMAG Anwendertreffen2018

© ika 2018 · All rights reserved15.11.2018Slide No. 31· 18jhe0005.pptx

Constraints: „Thermal Behaviour“

Exemplary Result:

89 kW PMSM

Buried magnets in V-shape

WLTC cycle

60 Deg. C coolant temperature

150mm

182mm

source: Grunditz, E. A.

Page 32: FEMAG Anwendertreffen2018

© ika 2018 · All rights reserved15.11.2018Slide No. 32· 18jhe0005.pptx

Agenda

Holistic Design of Electric Drive Modules

MATLAB & FEMAG

Calculations:

Parameter Identification

Losses

Cogging torque

Torque ripple

Calculated Contraints:

Mechanical load, Back EMF, Short circuit current

Demagnetization of PMs

Thermal behaviour

Outlook

Conclusion: FEMAG as a „Newbie“

Page 33: FEMAG Anwendertreffen2018

© ika 2018 · All rights reserved15.11.2018Slide No. 33· 18jhe0005.pptx

Outlook

Consideration of „Skin-“ and „Proximity Effect“ on phase resistance and copper losses

Makes it possible to find best winding technology for given machine and usage scenario

Loss minimization control

All data available already

Optimization of EM – Parameter studies

Reduction of calculation time

Page 34: FEMAG Anwendertreffen2018

© ika 2018 · All rights reserved15.11.2018Slide No. 34· 18jhe0005.pptx

Agenda

Holistic Design of Electric Drive Modules

MATLAB & FEMAG

Calculations:

Parameter Identification

Losses

Cogging torque

Torque ripple

Calculated Contraints:

Mechanical load, Back EMF, Short circuit current

Demagnetization of PMs

Thermal behaviour

Outlook

Conclusion: FEMAG as a „Newbie“

Page 35: FEMAG Anwendertreffen2018

© ika 2018 · All rights reserved15.11.2018Slide No. 35· 18jhe0005.pptx

FEMAG as a „newbie“

Uncommon DOS-like user interface

Handbook and online documentation lack some

information

Especially critical for my daily work with

students (bachelor- & master thesis)

Very accurate results

Faster than comparable FEM programs

Fast and friendly support

Page 36: FEMAG Anwendertreffen2018

© ika 2018 · All rights reserved15.11.2018Slide No. 36· 18jhe0005.pptx

Phone

Fax

Email

Internet www.ika.rwth-aachen.de

Institute for Automotive Engineering (ika)

RWTH Aachen University

Steinbachstr. 7

52074 Aachen

Germany

Contact

Jonas Hemsen M.Sc.

+49 241 80 25690

+49 241 80 22147

[email protected]

Page 37: FEMAG Anwendertreffen2018

© ika 2018 · All rights reserved15.11.2018Slide No. 37· 18jhe0005.pptx

References

GRUNDITZ, E. A.Design and assesment of battery electric vehicle powertrain, with respect to performance, energy consumption and electric motor thermal capabilityChalmers University of Technology, Göteborg, 2016

KACEM, HADJ MOEZThermal Modelling of Electric MachinesBookboon, Sfax, 2016

MÜLLER, G.; VOGT, K.; PONICK, B.Berechnung elektrischer MaschinenWiley, Somerset, 2012

KRISHNAN, R.Permanent magnet synchronous and brushless DC motor drivesCRC Press, Boca Raton, 2010

NATEGH, S.Thermal analysis and management of high-performance electrical machinesRoyal Institute of Technology (KTH), Stockholm, 2013

SCHRÖDER, D.Elektrische Antriebe – Regelung von AntriebssystemenSpringer-Verlag, Berlin, 2015

FINKEN, T.Fahrzyklusgerechte Auslegung von permanentmagneterregten Synchronmaschinen für Hybrid- und ElektrofahrzeugeShaker-Verlag, Aachen, 2011

WU, T.Lecture – Permanent Magnet Synchronous Motor (PMSM) IntroductionUniversity of Central Florida, Orlando, 2015