See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/308496607
Model-based design and testing for electric
vehicle driveability analysis
Conference Paper · June 2016
DOI: 10.1109/EEEIC.2016.7555884
CITATION
READS
1
32
6 authors, including:
Mathieu Sarrazin
Johan Gyselinck
Siemens industry software NV
Université Libre de Bruxelles
19 PUBLICATIONS 15 CITATIONS
157 PUBLICATIONS 877 CITATIONS
SEE PROFILE
SEE PROFILE
Herman Van der Auweraer
Claudia Martis
Siemens Industry Software NV
Universitatea Tehnica Cluj-Napoca
279 PUBLICATIONS 2,300 CITATIONS
81 PUBLICATIONS 244 CITATIONS
SEE PROFILE
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
DeMoTestEV – DEsign, Modelling and TESTing tools for Electrical Vehicles View project
ADEPT - ADvanced Electric Powertrain Technology View project
All in-text references underlined in blue are linked to publications on ResearchGate,
letting you access and read them immediately.
Available from: Sebastian Vasile Ciceo
Retrieved on: 14 November 2016
This material is presented to ensure timely dissemination of
scholarly and technical work. Copyright and all rights therein
are retained by authors or by other copyright holders. All
persons copying this information are expected to adhere to the
terms and constraints invoked by each author’s copyright. In
most cases, these works may not be reposted without the
explicit permission of the copyright holder.
© 2016 IEEE. Personal use of this material is permitted.
However, permission to reprint/republish this material for
advertising or promotional purposes or for creating new
collective works for resale or redistribution to servers or lists,
or to reuse any copyrighted component of this work in other
works must be obtained from the IEEE.
DOI: 10.1109/EEEIC.2016.7555884
URL: ieeexplore.ieee.org/document/7555884/
Sebastian V. Ciceo
Technical University of Cluj-Napoca
Str. Memorandumului nr. 28 Cluj-Napoca
Jud. Cluj, Romania
Postal Code: 400114
Phone: +4 0264 401 200
Fax: +4 0264 592 055
Email: sebastian.ciceo@mae.utcluj.ro
Model-Based Design and Testing for Electric Vehicle
Driveability Analysis
Sebastian Ciceo1, Yves Mollet2,3, Mathieu Sarrazin3, Johan Gyselinck2, Herman Van der Auweraer3 and Claudia
Marțiș1
1
Department of Electrical Engineering, Technical University, Cluj-Napoca, Memorandumului 28, Romania
2
Brussels School of Engineering, Université Libre de Bruxelles (ULB);
3
Siemens Industry Software NV
Abstract—In this paper a model-based design and testing
method focusing on the electric vehicle driveability aspect is
proposed. The design approach is divided into two steps. The first
step is the Model-in-the-Loop co-simulation coupling a vectorcontrolled electric drive modelled in MATLAB/Simulink to a
planar forward-facing electric vehicle LMS Imagine.Lab Amesim
model. The second step represents a mechanical-level Hardwarein-the-Loop test for a physical electric drive that integrates the
electric vehicle model in the real-time testing case. Two different
sampling times of the vehicle control unit are considered and
their influence on the vehicle responsiveness and on the
longitudinal jerk acting on the driver is analysed through both
offline simulation and real-time testing.
Keywords—Model-Based
Design;
Electric
Driveability; Hardware-in-the-loop; Model-in-the-loop
Vehicle;
I. INTRODUCTION
The growth of the hybrid and electric vehicle (EV) market
due to environmental concerns generates a need for rapid
development of ad hoc electromechanical components. To
achieve this objective, physics-based system engineering can
be merged with control engineering from a very early design
stage by means of virtual prototyping and testing [1,2].
Design techniques proposed in the literature combine signal
flow modelling for control design (signal simulation) with
physical modelling in various domains (e.g. the interaction
between electro-mechanical components) for system design
[3,4]. Tests can be conducted at signal, power and mechanical
levels in order to evaluate the controller, the combination of
controller and power electronics and the whole electrical drive
respectively [3,5].
The electromechanical system development and the control
system development adopt the ”V” approach that resumes to
propagate the system level requirements to component design
and validate the system performance at increasing interrogation
levels [1].
Because driveability is a subjective standard depending on
human-vehicle interaction, and therefore usually needing a real
person for evaluation, incorporating it in the design process
represents a cost-saving benefit by decreasing the time to
market of the designed product [6].
This research aims at acquiring a deeper insight in such EV
modelling and testing procedures by introducing the interaction
between the physical e-drive with the vehicle model during its
design process. The focus is on testing the effects of the
Vehicle Control Unit (VCU) control strategy on the driver‟s
comfort.
II. METHODOLOGY
The proposed approach enhances the closely related,
concurrently submitted work [7], where a Model-Based Design
(MBD) approach is adopted for testing the battery energy
consumption under different reference driving cycles on a
forward-facing EV model. A driver model taking the role of a
PI controller assesses the vehicle speed and gives acceleration
and braking commands to the VCU, consequently translating
these signals into electric motor torque commands and
mechanical braking.
Because the EV model has a gearbox with a fixed gear ratio
resulting in no gear shifting effect, the main component that
has an influence on the vehicle driveability is the VCU with its
control strategy.
By adding additional vehicle dynamics to the model (i.e.
longitudinal wheel slip, longitudinal stiffness and damping on
the suspensions) we can monitor and adjust the torque response
imposed by the VCU in order to keep the longitudinal jerk
acting on the driver within acceptable comfort limits.
The approach is divided in a MBD phase and a testing
phase of an EV system characterized by Model-in-the-Loop
(MiL) and by Hardware-in-the-Loop (HiL) implementation,
respectively (Fig. 1).
As noted in [1] MiL consists of combining multi-physics
simulation software with control software in an offline
environment, whereas in HiL hardware components are tested
while their environment is emulated in real-time on an
embedded platform. The MiL process starts prior to HiL testing
and can be used in a later development stage to further finetune the HiL testing in a safe and rapid fashion.
parameter
value
unit
stator and rotor inductance
0.052872
H
magnetizing inductance
0.051276
H
For driveability concerns we need to model the planar
motion of the vehicle (a model that has 3 degrees of freedom:
pitch rotation, longitudinal translation and vertical translations)
together with the suspension dynamics. The following
modelling assumptions are taken in consideration: the engine
block and longitudinal suspensions are considered blocked.
Because the parts through which the reaction force
generated by the force applied on the road through the tyre is
propagated have an influence on the vehicle‟s dynamic
behaviour [9], they have to be modelled as well (Fig. 2):
Flat road model - with the value of the adherence
equal to 1;
Wheels and tyres model - the simplified Pacejka
model with a dynamic method of longitudinal slip
calculation;
Planar vehicle suspension - having stiffness,
damping and unsprung mass as parameters;
Differential - ideal model, with no inertia or
friction.
Fig. 1. Diagram of the MiL and HiL co-simulation process
The MiL consists of a co-simulation approach where the
LMS Imagine.Lab EV and environment model is imported as
an interface block into a MATLAB/Simulink model of two
mechanically-coupled,
indirect-field-oriented-controlled
(IFOC) induction-machine (IM) drives. One machine is torque
controlled and represents the EV propulsion system while the
other one is speed controlled and represents the EV load (i.e.
the speed reference imposed by a roller in a real vehicle testbench). Simulink‟s fixed step solver clocked at the simulated
IGBT transistors switching frequency of 8 kHz (in order to
mimic the conditions of the real-time test-bench) runs the
whole co-simulation process.
LMS Imagine.Lab Amesim is an integrated simulation
platform for multi-domain mechatronic systems that is strongly
connected to the physical understanding of the phenomenon
described [8].
For the HiL stage, physical machines and inverters replace
their corresponding Simulink models. The test-bench consists
of two mechanically-coupled 45 kW, 8-pole IMs (extended
parameters are presented in TABLE I.) powered by two
modified FPGA-controlled inverters. The electric drive control
strategy is the same as in the MiL stage, however its
implementation on the test-bench takes into account the realtime requirements of rapid control prototyping. The LMS
Imagine.Lab Amesim model is integrated inside the control
through a process of real-time file generation for the target PC
running Xenomai, a Linux kernel real-time operating system.
TABLE I.
INDUCTION MACHINE PARAMETERS
parameter
value
unit
735
rpm
number of phases
3
-
moment of inertia
2.55997
kg m2
phase resistance
0.2718
Ω
rotor equivalent resistance
0.02474
Ω
stator leakage inductance
0.001595
H
rated Speed
Fig. 2. Vehicle suspensions, tyres and differential submodel in LMS
Imagine.Lab Amesim
III. SIMULATION AND EXPERIMENTAL RESULTS
As mentioned in [10], in order to achieve real-time
capabilities the offline model has to be reduced to the point
where the results converge to those of the real-time testing
scenario. This can be obtained with ease, in the MiL stage, by
using the Performance Analyzer tool inside LMS Imagine.Lab
Amesim. Its aim is to gather information about the model in
order to assess and improve the performance of the solving
process (e.g. the frequency at which the damping of the planar
vehicle suspension is varying inside the model can be
correlated to the integration step size in order to obtain accurate
real-time results).
Using two different values for the VCU control (see Fig. 3.)
sampling period, serving as the filtering time constant for the
torque command (0.1 and 0.2 s), the reaction of the VCU to the
acceleration command imposed by the driver in order to
produce the desired mechanical characteristics is investigated.
The sampling period can be characterized as the latency to
which the driver reacts to the test stimulus (i.e. the reference
speed profile).
(approximate 12 m/s3 at the beginning of the tip-in stage and 7 m/s3 when the acceleration stops in the online test).
Fig. 3. VCU torque and braking command algorithm in LMS Imagine.Lab
Amesim (pseudocode)
It can be noticed in the diagram that a maximum torque
control strategy is applied, where the acceleration command of
the driver (ranging from 0 to 1) is translated to a torque
command, where the pedal signal of 1 sets the electrical
machine torque command to the maximum motor torque
defined in a data file.
Simulation and experimental tests are conducted to assess
driveability and vehicular responsiveness and compare the MiL
and HiL results.
Because driveability is determined by the behaviour of the
powertrain especially in transient situations [6], the first tip-in
stage of the New European Driving Cycle (NEDC) where the
vehicle accelerates from 0 to 4.16 m/s in 4 s is taken as test
stimulus.
The monitored variables for the MiL and HiL are the IM
rotational speed (calculated in the offline model and measured
on the test-bench in real-time with an incremental encoder) IM
torque (calculated in the offline model and estimated in realtime based on the measured phase currents and shaft angle
position) describing the vehicle‟s responsiveness and
smoothness, and the estimated longitudinal jerk on the car body
describing the driving comfort.
The first value for the sample time constant taken into
consideration is 0.2 s. It can be noticed that, in both the offline
and real-time test cases, the driver reaction is too slow for the
initial change in the speed reference (the beginning of the tip-in
phase) leading to a mechanical torque overshoot in order to
compensate for the slow reaction. The obtained mechanical
parameters are not in line with the desired smoothness and
responsiveness for the electric machine and produce a
significant longitudinal jerk on the modelled car body
Fig. 4. Mechanical parameters and longitudinal jerk results for 0.2 s VCU
sampling time, in the MiL (a) and HiL (b) tip-in stage (first 18 s of the NEDC
cycle)
After reducing the sampling period to 0.1 s a decrease in
the longitudinal jerk to a reasonable value (approximate 7 m/s3
at the beginning of the tip-in stage and -4 m/s3 when the
acceleration stops in the online test) and the presence of a
smoother mechanical parameters response (a less visible torque
overshoot), in both MiL and HiL test scenarios can be noticed,
as exemplified in Fig. 4.
IV. CONCLUSION
As a conclusion, a method of estimating the driver comfort
and the EV driving response in order to meet comfort demands,
using an experimental approach based on calibrating the
sampling time of the VCU command algorithm, in MiL and
HiL EV design stages is proposed.
The co-simulation approach is preferred due to the
advantages of combining control engineering provided by
MATLAB/Simulink and system engineering provided by LMS
Imagine.Lab Amesim in the simulation stage, and the ease of
integrating both the control and system model in the HiL testbench for validation purpose.
ACKNOWLEDGMENT
The presented research was achieved in the context of the
research projects “FP7 ARMEVA” and “FP7 ASTERICS”.
REFERENCES
[1]
Fig. 5. Mechanical parameters and longitudinal jerk results for 0.1 s VCU
sampling time, in the MiL (a) and HiL (b) tip-in stage (first 18 s of the NEDC
cycle)
Another important remark is that by cross-checking the
results for the two tested VCU sampling times, from the MiL
simulation with the ones of the HiL test a good agreement
between the different variables can be noticed. This represents
a confirmation that electric vehicle comfort model behaves in
the same way in offline and real-time cases, thus having realtime capabilities The values of the longitudinal jerk and the
mechanical torque are slightly lower in the MiL simulation
because of the simplification made to the mechanical
parameters (e.g. ideal inertia, viscous forces) and the difference
in the complexity of the current PI controllers in the MiL
simulation and HiL testing.
H. Van der Auweraer, J. Anthonis, S. De Bruyne, and J. Leuridan,
“Virtual engineering at work: the challenges for designing mechatronic
products”, Engineering with Computers 29, no. 3 (2013): 389-408.
[2] E. Bringmann and A. Kramer, “Model-based testing of automotive
systems”, International Conference on Software Testing, Verification,
and Validation, 2008.
[3] K. Patil, S.K. Molla and T. Schulze “Hybrid vehicle model development
using ASM-AMESim-Simscape Co-Simulation for Real-Time HIL
applications”, SAE International, 2012.
[4] D.W. Gao and A. Emadi, “Modelling and simulation of electric and
hybrid vehicles”, Proceedings of the IEEE,Vol. 95, No. 4, April 2007.
[5] A. Bouscaryol, “Different types of Hardware-In-the-Loop simulation for
electric drives”, IEEE 978-1-4244-1666-0/08, 2008.
[6] A.F.A. Serrarens and F.E. Veldpaus, „„Driveability Assessment of a
CVT-Powertrain with Mechanical Torque Assist,‟‟ Proc. of the FISITA
2002 World Automotive Congress, Helsinki, Finland, 2002.
[7] S. Ciceo, Y. Mollet, M. Sarrazin, H. Van Der Auweraer and C. Marțiș,
“Model-Based Design and Testing for Electric Vehicle Energy
Consumption Analysis,” unpublished.
[8] J.C. Da Silva, “Virtual Environment for Dynamik Modelling of MultiDomain Systems”, Proceedings of the 18th COBEM, Ouro Preto (Brasil),
2005.
[9] I. Stoica, M. Bataus and M. Negrus, “Validation of a human-andhardware-in-theloop control algorithm”, U.P.B. Sci. Bull., Series D, Vol.
76, Iss. 4, 2014
[10] U. Schreiber, K. Todtermuschke and O. Back, “Experiencing and
definition of the real shifting confort at the virtual synchroniser in the
virtual powertrain”, 8th CTI Symposium, Innnovative Automotive
Transmissions, Berlin, 2009.