See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/335931962

Development and Control of a Camless Engine Valvetrain

Article in IFAC-PapersOnLine · January 2019

DOI: 10.1016/j.ifacol.2019.09.064

CITATIONS READS

5 5,111

5 authors, including:

Zlatina Dimitrova Tari Massinissa

École Polytechnique Fédérale de Lausanne SAFRAN GROUP
47 PUBLICATIONS 531 CITATIONS 15 PUBLICATIONS 70 CITATIONS

SEE PROFILE SEE PROFILE

Patrick Lanusse François Aioun

Bordeaux INP PSA Groupe
150 PUBLICATIONS 2,626 CITATIONS 70 PUBLICATIONS 584 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Zlatina Dimitrova on 29 January 2020.

The user has requested enhancement of the downloaded file.

 Mechatronics 57 (2019) 109–128

Contents lists available at ScienceDirect

Mechatronics
journal homepage: www.elsevier.com/locate/mechatronics

Robust control for an electromagnetic actuator for a camless engine, ☆,✰✰

Zlatina Dimitrova a,∗, Massinissa Tari a, Patrick Lanusse b, François Aioun b, Xavier Moreau c
a
PSA Groupe, Vélizy, France
b
Bordeaux INP, IMS CNRS UMR 5218, France
c
Université de Bordeaux, IMS CNRS UMR 5218, France

a r t i c l e i n f o a b s t r a c t

Keywords: The increase of the effective efficiency of the internal combustion engines is important for the fuel consumption
Electromagnetic actuator reduction and the CO2 emission reduction. The variable engine valve trains are important for the gas exchanges
Robust controller and influence directly the effective efficiency of the internal combustion engine. With the increased electrifica-
Engine variable valve train
tions of the vehicles, variable valve train based on electricity driven actuators presents interest for investigation
for gasoline engines for vehicular application.
This article presents an innovative application of an electromagnetic actuator for a future variable engine
valve train. The actuators are designed for small gasoline engine. The design of a high performing and dynamic
electromechanical system leads to a nonlinear behavior of the actuator. The nonlinearities that have to be faced
come from the magnetic laws and the mechanical friction. In the final control strategy, it is proposed to compen-
sate them both with nonlinear feedforward actions and with a linear robust feedback controller, which is able to
reject all what cannot be predicted. An experimental test bench is built for the correlation of the control model
with experimental measures. Control strategies are developed and the performance indicators of the actuators
are evaluated. The robust controller method uses CRONE system control methodology. The investigations show
that very good dynamic performances of the controller are obtained.

1. Introduction
This article presents an innovative concept of a Lorentz force bases
electromagnetic actuators for a full variable engine valve train, with
the associated control law. The method of the building of the control
law is also presented and the main performances of the actuator are
presented. The valve train is responsible for the gas exchange in the
engine. The variable profiles, regulating the opening and the closing of
the valves, help to improve the efficiency of the engine and are useful
for the internal combustion engines to reach the future emissions and
fuel consumption standards.
The camless valvetrain is considered to be a promising solution to
improve the engine performance. Most of the camless valvetrains, based
on reluctance force, suffer the problem of high impacts at valve seating,
which restricts its mass production.
The difficulty for the engine valves actuators in the case of Full Cam-
less engine is to cover the large range of forces and dynamics, required
for operating correctly all intake and all exhaust valves. Camless engine
means that such electromagnetic valves are deployed to the complete
intake and exhaust side. For previous studies and experiences with re-
luctance force based actuators, we know that main difficulties are:
• The precision of the actuators, where landing, the chock of the valves
inducing noise
• And the density of the delivered force
• The high electric consumption.
The literature review highlights the problems and the research done
on them. Lui and Chang in [1] focus on the valve seating performance of
an electromagnetic valvetrain (EMVT) with moving coil linear actuator.
The EMVT has inherent advantages to achieve low valve seating veloc-
ity. The seating performance of EMVT is evaluated by two indicators:
seating velocity and holding force. Mercorelli [2] studies a Lyapunov-
based adaptive control law for an electromagnetic actuator. A direct
adaptive non-linear control framework for a non-linear electromagnetic
actuator is presented in this paper. Starting from a Lyapunov approach
a continuous control law is calculated for the actuator. After that, an ap-
proximated sampled final control law is proposed. The resulting control
strategy is an adaptive one in which the main parameter depends on the
variable inductance of the actuator itself. Hoffmann and Stefanopoulou

☆
This work took place in the framework of the OpenLab “Electronics and Systems for Automotive" combining IMS laboratory and PSA Group company.
✰✰
This paper was recommended for publication by associate editor T.H. Lee.
∗
Corresponding author at: IPESE, EPFL SCI-STI-FM, Lausanne, Switzerland.
E-mail addresses: zlatina.dimitrova@mpsa.com (Z. Dimitrova), patrick.lanusse@ims-bordeaux.fr (P. Lanusse), francois.aioun@mpsa.com (F. Aioun),
xavier.moreau@u-bordeaux.fr (X. Moreau).

https://doi.org/10.1016/j.mechatronics.2018.12.004
Received 4 May 2018; Received in revised form 28 November 2018; Accepted 10 December 2018
0957-4158/© 2018 Elsevier Ltd. All rights reserved.
Z. Dimitrova, M. Tari and P. Lanusse et al.
Nomenclature
B magnetic field
I current amperage (A)
L length of the wire (m)
Uact voltage across the actuator (V)
Uv voltage across the actuator
m total mass of the moving part of the actuator (kg)
P weight of the actuator
Z (t) position of the valve (mm)
Zmes (t) measured position of the valve (mm)
Zref (t) reference position of the valve (mm)
Fact electromagnetic force delivered by the actuator
Ffrict friction force at the valve guide
Fpres pressure force exerted by the exhaust or intake gases
EMA Electro-Magnetic-Actuator
CSL control system law
CRONE French acronym which means fractional order robust
control
𝛽 0 (s) nominal open-loop
G0 (s) nominal plant of the actuator
G(s) uncertain plant of the actuator
CF (s) fractional controller
CR (s) rational controller
in [3] present the valve position tracking in electromechanical camless
valvetrains (EMCV). One of the main problems in the EMCV actuator is
the noise and wear associated with high contact velocities during the
closing and opening of the valve. The contact velocity of the actuator
and the valve can be reduced, and thus, noise and wear, by designing
a tracking controller that consists of a linear feedback and an iterative
learning controller. Peterson et al. in [4] describe a nonlinear self-tuning
control for soft landing of an electromechanical solenoid valve actuator
(reluctance force actuator), which is a kind passive actuator with springs
on the moving part. The authors studied low contact velocities or “soft
landing᾿ of the actuator on the solenoid faces and between the actua-
tor and the valve is also necessary in order to maintain similar noise
and wear levels with conventional camshaft-driven engines. Liu et al. in
[5] present an electromechanical valve actuator (EMVA) using solenoid
to actuate valve movement independently for the application of internal
combustion engine. They proposed an EMVA structure by incorporating
the hybrid magneto-motive force (MMF) implementation in which the
magnetic flux is combined by the coil excitation and permanent mag-
nets. This type is also a passive type of actuator. Making use of the
dedicated flux arrangement, the proposed device can be used to fulfill
the VVT features with reduced power source requirement and less elec-
tric device components. A dual flux channels EMVA is detailed and the
design procedures are presented. Bolat and Sivrioglu in [6,7] present re-
search related to electromagnetic actuators and robust control method
applied to a small-scale wind turbine blade.
The variable engine valve train plays an important role to control
the gas exchange and constitutes an opportunity for the optimization
of the combustion process [8–10]. The components of the engine valve
train are exposed on important efforts, coming from the gases under
high pressure in the combustion chamber, especially when the exhaust
valve has to open to evacuate the burned gases. One can distinguish a
family of variable valve trains called mechanical variable valve train.
An example of a mechanical variable valve train is given in Fig. 1.
This kind of valve train the lift profile is defined by the camshaft profile
and the lift is managed by a mechanical system. This system is associated
with a “camephaser", which allows phasing the times for opening and
closing of the valve. A typical lift profile for the mechanical valve train
system is presented in Fig. 2.
110
Mechatronics 57 (2019) 109–128
Fig. 1. Mechanical variable valve train [11].
A second family of variable valve trains are the electromagnetic ac-
tuators for valves. The valve trains composed by electromagnetic actua-
tors do not contain camshaft and for this reason the engines and called
“camless engines”. The “Camless technology” allows to drive individu-
ally on each valve and so to find the optimal operating conditions of
the engine for high engine efficiency. This technology allows also some
advanced functions of the valve train as cylinder deactivation [12] or
the pneumatic hybridization of the engine, where the electromagnetic
valves manage the compressed air flows between the engine and the
compressed air tank [13]. The technology to act the valves could be
hydraulic, electro mechanic and electromagnetic. The hydraulic actu-
ators for the Camless technology are presented in [8,10,14,15]. They
are based on the application of oil under pressure and piezo-valve that
act on the engine valve. The performances of the hydraulic actuators
depend on the external temperature and its relation with the oil viscos-
ity. The electro mechanic actuators use a brushless electric motor and a
mechanical system using spring to act the valves [14]. This technology
is difficult to be integrated in the vehicles because of its lack of com-
pactness. The third category of Camless engine is using electromagnetic
actuators to act the engine valves [15–17].
Fig. 3 presents one of the most used concepts of electromagnetic
actuators.
The electromagnetic actuator (EMA) is composed of two springs, in
opposition, two electromagnets and an armature located between them.
For such actuators the moving masses are rather important and increase
their time of response.
An electromagnetic actuator with moving magnet is presented in
[18]. This paper develops a miniature electromagnetic energy harvester,
which can absorb gas pressure fluctuations in industrial piping or natu-
ral environment and then supply energy to electronic devices. The mov-
able magnet in the middle of the housing is utilized as a piston on which
the gas pressure directly acts.
The contribution of this article is the use of an innovative concept
of engine valve train based on Lorentz force electromagnetic actuator,
which is characterized by performant and efficient magnetic circuit,
lightweight moving part, direct driven by the current. This is an ac-
tive actuator, which solves the problems highlighted in the previous re-
searched – the control robustness and precision. A method for control,
based on simulation and experimental steps is presented for the inno-
vative actuator. The nonlinear behavior of the actuator results from the
mechanical friction and the magnetic circuit. Furthermore, the physical
parameters of the designed model are uncertain. The developed control
strategy proposed to compensate uncertainties and nonlinearities both
with linear or nonlinear feedforward actions and with a linear robust
Z. Dimitrova, M. Tari and P. Lanusse et al.
feedback controller, which rejects all what cannot be predicted. The per-
formances of the control are assessed. The application of the controlled
actuator is for a gasoline engine with camless valvetrain on the intake
and the exhaust side.
2. Electromagnetic actuator for camless engine
2.1. System description
The Lorentz force based actuator module presented in this article is
designed for two valves, to act on an individual way each of the valves.
The modules are equivalent for the intake or the exhaust side of the
engine. The principle and the details of the actuator are described in
[19].
The actuator module Fig. 4) is composed of one fixed part that forms
the magnetic circuit, including one permanent magnet and ferromag-
netic parts. The dimensions of the actuator are 80 mm of length, 55 mm
of width and 50 mm of height. The mass of the total actuator is 440 g.
Fig. A1 presents schematically the innovative actuator. The permanent
magnet (22, Fig. A1) is located in the middle of two ferromagnetic parts
(23a and 23b, Fig. A1). The magnet is NdFeB. The ferromagnetic mate-
rial is bronze. The magnetic circuit is closed by using full iron cylinders
(26 and 24). The part 28 is also ferromagnetic and is used as cover to
111
Mechatronics 57 (2019) 109–128
Fig. 2. Lift profile for a mechanical valve train and an engine
speed of 6000 rpm.
Fig. 3. Example of electromagnetic actuators, passive
types [11].
maintain fixed the cylinders (26 and 24). They are crossing the sand-
wich of the permanent magnet ((22) and the two ferromagnetic parts
(23a and 23b). The set of part indicated as (19) in Fig. A1 is fixed. The
moving part of the actuator are indicated in Fig. A1, as sets of parts (21)
and (20). The actuator has one moving part for every valve. The mov-
ing part is composed of one carcass (42) and (36), which are concentric
with the iron cylinders (26) and (24). The carcass is wearing the wires
(43, 44, 34, 35) and the valve is fixed on its bottom extremity of the
carcasses (42) and (36). The material of the carcasses is chosen to be
a composite (PEEK). This allows enough good mechanical properties to
pass the efforts for lightweight characteristics. The order of magnitude
of the moving mass is around 36 g, which insure low kinetic energy de-
mand while moving. The composite is neutral for the electromagnetic
field. The wires of every moving part are connected through flexible con-
nections to the vehicle electric grid and can operate on different voltage
levels. Other advantage of the innovative design is the small air gap
in the magnetic circuit, important for the electric consumption of the
actuator. The electric connection is independent for each valve, which
insures an individual control and acting of the valves.
Fig. A2 illustrates the position of the actuator when the valves are
open. Fig. A3 illustrates the position of the actuator when the valves are
closed. The maximal valve lift is 7 mm.
Z. Dimitrova, M. Tari and P. Lanusse et al.
Fig. 4. Innovative electromagnetic actuator design.
The size of the permanent magnet and the shape of the ferromag-
netic parts that surround the permanent magnet are studied and op-
timized to deliver the target of the maximal force. The ferromagnetic
cylinders close the magnetic circuit with a very small air gap, which
is needed for good efficiency of the actuator. The efficiency is directly
related with the electricity consumption of the actuators, which has to
be minimized. The discussed actuator design is innovative its magnetic
circuit and lightweight moving part, including the valves and the car-
casses, where the copper is wired.
The volume of the electromagnetic actuator module is compact. The
dimensions are 50 mm∗ 80 mm∗ 55 mm. The actuator is delivering an
electromagnetic force and is governed by:
𝐹𝑎𝑐𝑡 (𝐼, 𝑍 ) = 𝐵 (𝑍 ) 𝐼 𝐿𝑐𝑜𝑖𝑙 , (1)
where Fact is the electromagnetic force delivered by the actuator in [N],
B is the magnetic field in [T], I is the current amperage [A], Lcoil is
the length of the wire in [m] and Z is the valve position (m). The size
of the permanent magnet is related to the magnetic field and then to
the electromagnetic force, delivered by the actuator and governed by
Eq. (1) The relation between magnetic field, current and valve position
has been modeled (Fig. 5). This model has been obtained after simula-
tion with Finite Elements Method Magnetics (FEMM), which contains
Maxwell equations. The FEMM model of the actuator is validated with
experimental results of the prototypes, in static conditions, for a fixed
position of the valve. Fig. 6 presents the correlation of the FEMM model
and the measures of the actuator. As the current limit is 90 A, the elec-
tromagnetic maximal force delivered by the actuator on each valve is
400 N. This level of force allows deploying the module also on the ex-
haust side, where the effort to open the combustion chamber under pres-
sure is important. The actuator does not contain a mechanical spring
with an energy storage function. Its structure is active, without a pas-
sive mass/spring system and present main advantages for the controlling
of the actuator. The active principle of the controlling avoids the passive
resonance modes, which are difficult to control and provoke vibrations
and noise during when the valve lands on the valve seat.
The introduced actuator has an efficient magnetic circuit. Due to his
design the actuator is directly driven by the current and is belongs to
the so-called family of active actuators. This fact introduces the novelty
of the study and also presents an advantage for the active control of the
actuator and thus the precision obtained by the control of the actuator.
The method used for the control of the active actuator presents interest
for the state of the technology. The implementation and the experimen-
tal results are as well useful.
112
Mechatronics 57 (2019) 109–128
2.2. Actuator nonlinear model
The model determination of the actuator requires the study of its
mechanical part and its electrical part. Thus, the nonlinear model of the
actuator consists of 3 interconnected sub-parts (Fig. 7), one relating to
the electrical part of the system, one of its electromechanical part and
one relative of its mechanical part.
The electrical part allows the taking into account (modeling) of the
electric circuit of the actuator. From the voltage across the actuator
(Uact (t)) and the electromotive force of the actuator (voltage Uv (t)), it is
possible to express the current I(t) passing through the actuator.
The Electromechanical Part models the connection between the elec-
trical and mechanical part of the actuator. By considering the limited
operating domain of the valve, the model of the force produced by the
electromagnetic actuator defined by (1) can be reduced to a fifth order
polynomial. The Mechanical Part allows a simplified modeling of the
mechanical part of the actuator (Fig. 8).
Newton’s second law allows taking the balance of the forces having
an influence on the movement of the valve:
𝑑 2 𝑍 (𝑡)
𝐹𝑎𝑐𝑡 (𝑡) − 𝐹𝑓 𝑟𝑖𝑐𝑡 (𝑡) − 𝐹𝑝𝑟𝑒𝑠 (𝑡) − 𝐹𝑣𝑖𝑠𝑐 (𝑡) + 𝑚𝑔 = 𝑚 ⋅ (2)
𝑑 𝑡2
where m is the total mass of the moving part of the actuator, mg is its
weight, Z(t) is the position of the valve, Fact (t) is the force generated
by the actuator, Ffrict (t) is the friction force at the valve guide, Fvisc (t) is
viscous force and Fpres(t) is the force exerted by the pressure difference
between the gases in the cylinder and the exhaust or intake gases. The
actuator force depends on the current amperage. The direction of the
current defines the sign of the force of the actuator. When the current is
positive (I > 0), the actuator force is also positive (Fact > 0) and the en-
gine valve opens. The negative current (I < 0) procures a negative force
(Fact < 0) and the valve closes.
2.3. Experimental testing bench
For an easy study of the actuator and rapid implementation and val-
idation of the control law, which is to be designed, a dSpace proto-
typing electronic unit has been chosen. Fig. 9 presents the block dia-
gram of the actuator control set-up that is implemented. Taking into ac-
count several aspects which affect the experimental set-up, a sampling
time Ts = 50 μs is chosen. The control software receives the position
sensor voltage through and generates current commands for the power-
electronic unit which converts the requested current to a control voltage
Uact (t) which is applied to the actuator by means of a PWM technique
with 20 kHz frequency. Due the new vehicle battery specification limit,
the magnitude of the control voltage is limited to 48 V and that of the
current to 90 A.
3. Control law of the system (CSL)
In order to make the EMA a reliable solution to perform Variable
Valve Timing operation in Camless engines, an extremely delicate stage
to be addressed is the control law system design [20–23]. The main
function of the CSL is to move the valve from the closed to the opened
position (and vice versa) while satisfying robust tracking of the reference
position valve [24], low energy consumption requirements and avoiding
high intensity impacts [24] (ex: impacting velocity less than 0.4 m/s at
6000 rpm). For this purpose, a control law architecture presented by
Fig. 10 is designed. It consists of a robust controller synthesized with
the CRONE methodology [25–31] allowing a control law by feedback
position provided by a sensor system and a complementary control with
a position feedforward.
3.1. Frequency domain system identification of the actuator
The global model of the actuator is highly nonlinear, and because
of the natural integral behavior, the limits of displacement, the vari-
Z. Dimitrova, M. Tari and P. Lanusse et al. Mechatronics 57 (2019) 109–128

Fig. 5. Electromagnetic modelling of the actuator.

Fig. 6. Characteristics of the actuator: FEMM model ver-

sus measurements.

able friction force and the very high speed of the actuator, it is diffi- Based on the experimental frequency response, the input-output dy-
cult to achieve an open loop and time-domain classical system iden- namical behavior of the actuator is modelled by a black-box model pa-
tification of its dynamical behavior. Thus, for the design of the feed- rameterized as follows:
back controller with the frequency-domain based CRONE methodology, ( )
a frequency-domain system identification of the linear behavior of the 𝐾0 1 + 𝑠∕𝜔
𝑝0
𝐺 0 (𝑠 ) = ( ( )2 )( ( )2 ) . (3)
actuator around an operating point close to the middle valve position is 2𝜉 𝑠 2𝜉 𝑠
achieved. 𝑠 1 + 𝜔 10 + 𝜔 𝑠 1 + 𝜔 20 + 𝜔 𝑠
𝑛10 𝑛10 𝑛20 𝑛20
First, for the local stabilization of the valve position around 3 mm,
a very simple PID controller is designed from the gain and phase of the The parameters of the transfer function (3) are optimized by mini-
system roughly measured in open loop at 1000 rad/s. The mean value mizing the following cost function:
of uact and its sine variation amplitude have been defined by a trial and ∑ ( )2 ( ( ) ( ))2
| |
error process. The PID controller ensures a 1000 rad/s open-loop gain |𝐺sys (𝜔)| − ||𝐺0 (j𝜔)||dB + arg 𝐺sys (𝜔) − arg 𝐺0 (j𝜔) .
| |dB
crossover frequency and a 40° phase margin. Using the control-system 𝜔∈[1000,10000]

law presented by Fig. 9, the frequency response of the system is mea- (4)
sured in closed-loop: the PID controller is then used, the reference valve
The parameter optimization provides:
position zref is defined by a mean value of 3 mm and a sine variation
with a frequency successively from 1000 rad/s to 10,000 rad/s, no feed- 𝐾0 = 276; 𝜔𝑝0 = 9000𝑟𝑎𝑑∕𝑠;
(5)
forward (uFF = 0) and filter effects are used. The experimental gain and 𝜔𝑛10 = 1597𝑟𝑎𝑑∕𝑠; 𝜉10 = 0.8446; 𝜔𝑛20 = 9840𝑟𝑎𝑑∕𝑠; 𝜉20 = 0.1069.
phase measures |𝐺sys (𝜔)|dB and arg(𝐺sys (𝜔)) are obtained by comparing
the sine part of zmes to that of uact . The Bode diagram of Gsys is pre- Fig. 11 permits the comparison between the experimental frequency
sented on Fig. 10. response and that of the linear model (3). The agreement of these fre-
quency responses validates the proposed model.

113
Z. Dimitrova, M. Tari and P. Lanusse et al. Mechatronics 57 (2019) 109–128

Fig. 7. Block diagram of the nonlinear model of the actuator.

for a robust controller design. It is defined by:

( )
𝑍𝑚𝑒𝑠 (𝑠) 𝐾 1 + 𝑠∕𝜔
𝑝
𝐺(𝑠) = = ( ( )2 )( ( )2 ) , (6)
𝑈𝑎𝑐𝑡 (𝑠) 2𝜉 𝑠 2𝜉 𝑠
𝑠 1 + 𝜔 1 + 𝜔𝑠 1 + 𝜔 2 + 𝜔𝑠
𝑛1 𝑛1 𝑛2 𝑛2

with
[ ]
⎧𝐾 ∈ 𝐾 0 , 𝐾 ∗ 1 . 5
⎪ [1 . 5 0
]
⎪𝜔 ∈ 𝜔 𝑝 0 , 𝜔 ∗ 1 . 1
⎪ 𝑝 [1.1 𝑝0
]
Fig. 8. Mechanical part of the actuator. ⎪𝜔 ∈ 𝜔𝑛10 , 𝜔 ∗ 1.1
⎨ 𝑛 1 . 𝑛 10 (7)
[ 1 1 ]
⎪𝜔 ∈ 𝜔𝑛20 , 𝜔 ∗ 1.1
⎪ 𝑛2 [ 1.1 𝑛20
]
⎪𝜉1 ∈ 𝜉10 − 0.1, 𝜉10 + 0.1
The nominal black-box model (3) and its parameters defined by ⎪ [ ]
⎩𝜉2 ∈ 𝜉20 − 0.05, 𝜉20 + 0.05
(5) are derived to define an uncertain plant model that will be used

Fig. 9. Block diagram of the actuator control

implementation set-up.

114
Z. Dimitrova, M. Tari and P. Lanusse et al. Mechatronics 57 (2019) 109–128

Fig. 10. Structure of the control system law

(CSL).

Fig. 11. Experimental frequency response (_ _ ); linear model frequency response (__ ).

Uncertainty ranges are defined by a rule of thumb, which is com- the nonlinear model described by Section 2.2. The optimized mechanical
monly used by the industrial partner of the project. Fig. 12 presents the and electrical parameter values are:
set of frequency responses of G that defines the uncertain plant. It shows
⎧𝐿𝑏𝑜𝑏 = 95.067 μ𝐻
significant uncertainties both on the gain and on the phase of the plant. ⎪𝑅 = 0.03050 Ω
From industrial point of view, the controller has to guaranty the ro- ⎪ 𝑖𝑛𝑖𝑡
⎨𝑚 = 0.0676 𝑘𝑔 (9)
bustness with respect to these uncertainty intervals. ⎪𝐿𝑐𝑜𝑖𝑙 = 7.4409 𝑚
Even if the linear model of the uncertain plant is useful to design a ⎪
⎩𝐾𝑣𝑖𝑠𝑐 = 1.5 𝑁𝑠∕𝑚
controller whose robustness could be verified by using the experimen-
tal system, the nonlinear model presented by Fig. 7 is needed to assess Figs. 13 and 14 permit the comparison between the experimental
the control law (feedback and feedforward controller) for more realistic frequency and time responses (for an engine speed of 6000 rpm) and
working cases, for instance when pressure forces are taken into account those of the nonlinear model. The agreement of these responses validates
as disturbances. Thus, the experimental frequency response is used to the proposed model.
optimize the mechanical, electrical and electromechanical parameters
of the nonlinear model. The cost function to be minimized is defined 4. Robust CRONE control design
by:
4.1. CRONE control-system design (CSD) methodology
∑ ( )2 ( ( ) ( ))2
| |
|𝐺sys (𝜔)| − ||𝐺NL (j𝜔)||dB + arg 𝐺sys (𝜔) − arg 𝐺NL (j𝜔) ,
| |dB The CRONE (a French acronym that means fractional order robust
𝜔∈[1000,10000]
control) CSD methodology is a frequency-domain approach developed
(8)
since the eighties [25–29,31]. It uses the common unity-feedback config-
where |𝐺NL (j𝜔)|dB and arg(𝐺NL (j𝜔)) are gain and phase of the numerically uration presented by Fig. 8 and the design of the robust controller can be
linearized model (obtained with the linmod Matlab-Simulink function) of based on the uncertain frequency response of the perturbed plant only,

115
Z. Dimitrova, M. Tari and P. Lanusse et al. Mechatronics 57 (2019) 109–128

Fig. 12. Bode diagram of nominal plant frequency response (__ ) and set of frequency responses of G(s) (_ _ ).

Fig. 13. Experimental frequency response (__ ); nonlinear model frequency response (__ ).

116
Z. Dimitrova, M. Tari and P. Lanusse et al. Mechatronics 57 (2019) 109–128

Fig. 14. Experimental time responses (__ ); nonlinear model time responses (__ ), a) position, b) voltage, c) current.

for instance provided by a black-box model. Three CRONE CSD meth- steady state errors:
ods have been developed. Each of them denotes a generation of CRONE 𝑠 𝜇
design. The first two are based on real fractional order differentiator or
( 𝜔 )𝑛I ⎛⎜ 1 + 𝜔1
⎞
⎟ ( 1 √
𝐶 (𝑠 ) = 𝐶 0 1 + I 𝑠 )𝑛F with 𝜔cg = 𝜔1 𝜔2 . (10)
integrator and provide robust feedback controller for plants with gain 𝑠 ⎜1 + ⎟ 𝑠
⎝ 𝜔2 ⎠ 1+ 𝜔
F
variations.
The first generation CRONE controller is defined within a fre- Such a transfer function defines a cascade fractional order PID𝜇 (F)
quency range [𝜔A , 𝜔B ] around the desired open-loop gain crossover controller: integer order nI defines its number of integrators; integer or-
frequency 𝜔cg from the transfer function of a fractional order n integro- der nF defines its high-frequency roll-off; [𝜔1 , 𝜔2 ] covers the frequency
differentiator. The CRONE controller includes a band-limited integrator range where 𝜔cg varies at time of system gain variations; corner frequen-
and a low-pass filter (Fig. 15) to manage the control effort level and cies 𝜔I , 𝜔h are respectively set lower and greater than 𝜔1 and 𝜔2 ; finally
fractional order 𝜇 and gain C0 are computed to ensure a nominal phase

117
Z. Dimitrova, M. Tari and P. Lanusse et al. Mechatronics 57 (2019) 109–128

related to tracking, regulation and control effort level, 5 inequality con-

straints are to be fulfilled for all plants (or parametric states of the plant):

inf |𝑇 (j𝜔)| ≥ 𝑇l (𝜔), sup |𝑇 (j𝜔)| ≤ 𝑇u (𝜔), sup |𝑆 (j𝜔)| ≤ 𝑆u (𝜔)

𝐺 𝐺 𝐺 (16)
sup |𝐶𝑆 (j𝜔)| ≤ 𝐶 𝑆u (𝜔) and sup |𝑆𝐺(j𝜔)| ≤ 𝑆 𝐺u (𝜔)
𝐺 𝐺

with
{ 𝐺(𝑠)𝐶 (𝑠) 1
𝑇 (𝑠) = 1+𝐺(𝑠)𝐶 (𝑠)
𝑆 (𝑠 ) = 1+𝐺(𝑠)𝐶 (𝑠)
(17)
𝐶𝑆 (𝑠) = 1+𝐺𝐶((𝑠𝑠))𝐶 (𝑠) 𝑆𝐺(𝑠) = 1+𝐺𝐺((𝑠𝑠)𝐶
)
(𝑠)

The optimal open-loop parameters position the open-loop frequency

Fig. 15. Frequency response of a cascade fractional order PID𝜇 (F) controller
for: nI = 2, nF = 1 and 𝜇 = 0.75.
margin and gain crossover frequency. Even if the system gain varies, the
constant controller phase 𝜇𝜋/2 around 𝜔cg ensures the constancy of the
phase margin when the plant phase is constant both versus frequency
and plant parameters. The achievable rational version of this controller
can be obtained using the Oustaloup’s approximation [25,26,31].
The second CRONE CSD generation can be used even if the plant
phase varies versus the frequency (but is not uncertain) around 𝜔cg . It
uses a nominal open-loop transfer function that includes a band-limited
fractional order integrator. The third generation extends the field of ap-
plication of the second one and should be used for any kind of plant
perturbation model. The nominal open loop (for the nominal plant G0 )
is defined by:
𝛽0 (𝑠) = 𝛽l (𝑠)𝛽m (𝑠)𝛽h (𝑠)𝛽G (𝑠). (11)
uncertainty domains defined by possible values of G(j𝜔)/G0 (j𝜔) in order
to not overlap the low stability margin areas of the Nichols chart. The
parameterization of the nominal open-loop transfer function by com-
plex fractional orders simplifies the optimization considerably. During
optimization, a complex order has the same function as a whole set of
parameters found in common rational controllers. When it is useful, N−
and N+ are different from 0 to increase the number of tuning parame-
ters. After the optimization of the open-loop parameters, the fractional
controller CF (s) is defined by its frequency response:
𝛽0 (j𝜔)
𝐶F (j𝜔) = . (18)
𝐺0 (j𝜔)
This desired frequency response is fitted by a low order rational
transfer function CR (s), which can be used to control the model of the
system and then the experimental system itself.
The CRONE methodology has already been compared to other tech-
niques and always leads to very good performances as it takes plant
perturbations without any conservatism. Related to the control problem
tackled in [27], Landau et al. [32] present a comparison of many design
methodologies (H∞, QFT, robust pole placement, GPC, etc.) for a SISO
perturbed and lightly damped system. QFT and CRONE controllers were
the best.
4.2. CRONE approach for the electromagnetic actuator control

‒ 𝛽 m (s) is a set of band-limited complex fractional order integrators: The designed controller has to be discrete-time implemented for an
𝑁 + implementation purpose within a dSpace system (or within processor in
∏
𝛽 m (𝑠 ) = 𝛽m 𝑘 (𝑠), (12) the general case). So, in order to take into account the effect of time-
𝑘=−𝑁 − discretization, the initial digital control system design problem with the
with: sample period Ts is transformed into a pseudo-continuous problem by:

( )𝑎𝑘 ( {( )i𝑏𝑘 })−𝑞𝑘 sign(𝑏𝑘 ) • Taking into account the transfer function of a zero-order hold on the
sign(𝑏𝑘 )
1+𝑠∕𝜔 1+𝑠∕𝜔
𝛽 m 𝑘 (𝑠 ) = 𝐶 𝑘 𝛼𝑘 1+𝑠∕𝜔𝑘+1 ℜ𝑒∕i 𝛼𝑘 1+𝑠∕𝜔𝑘+1 plant input;
𝑘 𝑘
( ( )2 ( )2 )1∕2 • Computing the z-transform of the continuous-time set {plant and
( )1∕2 𝜔r 𝜔
𝛼𝑘 = 𝜔𝑘+1 ∕𝜔𝑘 for 𝑘 ≠ 0 and 𝛼0 = 1 + 𝜔 ∕1 + 𝜔 r holder};
0 1
• Achieving a bilinear variable change, for instance
(13) z−1 = (1−w)/(1+w).
‒ 𝛽 l (s) is an integer order nl proportional integrator and 𝛽 h (s) is a low- For an engine speed of 6000 rpm, Fig. 2 shows that the lift profile
pass filter of integer order nh : lasts less than 8 ms. To describe it well, the sample period is defined
(𝜔 − )𝑛l ( )−𝑛h
𝑠 by Ts = 50 μs. So, G(s) becomes successively GH (s) = ZOH(s)G(s), where
𝛽l (𝑠) = 𝐶l −𝑁 + 1 , 𝛽h (𝑠) = 𝐶h +1 . (14) ZOH(s) is the transfer function of the zero-order hold
𝑠 𝜔𝑁 +
‒ 𝛽 G (s) is a transfer function defined from the right half plane or un- 1 − 𝑒−𝑠𝑇𝑠
𝑍𝑂𝐻 (𝑠) = (19)
derdamped poles and zeros of the nominal model of G0 (s). 𝑠
then G(z) = Z{GH (s)} and finally G(w). Fig. 16 shows the set of pseudo-
Gain K𝛽 (product of all Ck , Cl and Ch ) is such that 𝜔r is the nominal
continuous frequency responses of G(s). The controller C(w) can then
closed-loop resonant frequency. Order nl has to be set to manage the
be designed using a frequency-domain method knowing that G(w) (with
accuracy provided by the feedback control-system. Order nh has to be
w = jv and the normalized pseudo-frequency v = tan(𝜔Ts /2)) is the exact
set to obtain a bi-proper or strictly-proper controller.
frequency response of G(z) (with z = ej 𝜔T s ).
When N+ = N− = 0, only four open loop parameters are optimized in
Fig. 17 shows the Nichols plot of G(w) whose uncertainty domains
order to minimize a robustness cost function J:
take into account the uncertain parameters (7) of the plant model.
𝐽 = 𝑠up ||𝑀T || − 𝑀T 0 , (15) By taking into account the frequency domain uncertainty,
𝐺 𝛽 0 (w) = CF (w)G0 (w) is tuned. In this application case, a controller with
where MT0 is a required value of the resonant peak of the nominal com- a high static gain (without integrator) is enough to ensure a good track-
plementary sensitivity function T0 (s). To manage precisely performance ing of a steady state reference signal with a good rejection of the effect

118
Z. Dimitrova, M. Tari and P. Lanusse et al. Mechatronics 57 (2019) 109–128

Fig. 16. Nominal (_ ) and extreme (_ and _ )

pseudo-continuous frequency response of G(s).

Fig. 17. Nichols plot G: nominal frequency response (__ ); un-

certainty domains (__ ).

of disturbance on the plant input. Thus, the low frequency order nl of N+ = N− = 0 seems enough to solve the optimization problem. The nom-
𝛽 l (w) is set to 1. Three right half plane zeroes of G0 (w) define 𝛽 G (w). inal open loop transfer function to be tuned is thus defined by:
Thus, to limit the control effort sensitivity at high frequency and to ob- (𝑣 )1 ( 1 + 𝑤∕𝑣 )𝑎0
tain a strictly proper controller, the high frequency order nh of 𝛽 h (w) 𝛽 0 (𝑤 ) = 𝐾 β 0 + 1 1
𝑤 1 + 𝑤∕𝑣0
needs to be 4. To track reference valve position for engine speed up to ( )
( {( ) })−𝑞0 sign 𝑏q0
6000 rpm, the desired closed-loop bandwidth should be at least more 1 + 𝑤∕𝑣1 i𝑏0
or less 900 rad/s, the corresponding pseudo-frequency is 0.02. The re- × ℜ𝑒∕i 𝛼0
1 + 𝑤∕𝑣0
quired nominal resonant peak MT0 is 1 dB, the resonant peak limitation ( )( )( )
(T) is 4.5 dB, the sensitivity function limitation (S) is 6 dB and the plant 1 − 𝑤1 1 − 1𝑤 .22
𝑤
1 − 95
input sensitivity function limitation (SG) is 3 dB. Taking into account the × ( )4 (20)
𝑤
measurement noise, the limitation of the control effort sensitivity func- 𝑣
+ 1
1
tion (CS) is 48 dB. Fig. 19 shows theses frequency domain constraints.

119
Z. Dimitrova, M. Tari and P. Lanusse et al. Mechatronics 57 (2019) 109–128

Fig. 18. Nichols plot of 𝛽 0 nominal frequency response (__ ); uncertainty domains (__ ); 1 dB M-contour related to MT0 (__ ).

Fig. 19. Frequency-domain constraints (—),

nominal and extreme perturbed values of
|T(j𝜔)|, |S(j𝜔)|, |CS(j𝜔)| and |SG(j𝜔)|.

The 4 optimized parameter values for 𝛽 0 are: Yr = │𝛽 0 (j □r )│ = 6 dB;
vr = 0.015; v0 = 0.018; v1 = 0.1. Thus, a0 = 0.99, b0 = 0.81, q0 = 3 and
K𝛽 = 2.28. Fig. 18 presents the Nichols plot of the optimized open loop
frequency response 𝛽 0 (s). The final value of the cost function J is 1.07 dB
and all sensitivity constraints are met (Fig. 18): the greatest values of the
resonant peak of T, S, SG and CS are respectively 2.07 dB, 5 dB, −10 dB
and 47 dB. This last value is very close to the 48 dB limit which shows
that a larger bandwidth cannot be obtained: depending on the paramet-
ric state of the system, the close loop bandwidth is more or less between
1000 rad/s and 5000 rad/s.

120
Z. Dimitrova, M. Tari and P. Lanusse et al. Mechatronics 57 (2019) 109–128

Fig. 20. Opening and closing valve at 1000rpm: refer-

ence signal (__ ); experimental response (__ ).

Fig. 21. Control voltage at 1000 rpm.

By using relation (18), the frequency response of the fractional ver-

sion CF (s) of the controller is fitted by an order 7 rational transfer func-
tion:
4.887𝑒07 𝑠6 + 1.247𝑒07 𝑠5 + 4.18𝑒06 𝑠4 + 4.428𝑒05 𝑠3 + 2.064𝑒04 𝑠2 + 515.3 𝑠 + 6.395
𝐶 (𝑤 ) = (21)
1.149𝑒05 𝑠7 + 2.594𝑒05 𝑠6 + 2.01𝑒05 𝑠5 + 7.881𝑒04 𝑠4 + 1.689𝑒04 𝑠3 + 1858 𝑠2 + 88.32 𝑠 + 1

This fitting leads to an approximation error. Nevertheless, the order

of the controller has been chosen such that the error is small enough
(less than 1° and 0.1 dB for phase and magnitude) in order not to impair
significantly the optimization of the closed-loop robustness cost function
(15) subject to the performance constraints defined by (16).
Finally, the digital controller C(z) is obtained by achieving the in-
verse variable change
w = (1−z−1 )/(1+z−1 ) of C(w):
98.05 𝑧7 − 424.2 𝑧6 + 651.5 𝑧5 − 253.9 𝑧4 − 435.3 𝑧3 + 618.4 𝑧2 − 314.3 𝑧 + 59.67
𝐶 (𝑧 ) = (22)
𝑧7 − 4.102 𝑧6 + 6.932 𝑧5 − 6.15 𝑧4 + 2.97 𝑧3 − 0.6913 𝑧2 + 0.02987 𝑧 + 0.01062

121
Z. Dimitrova, M. Tari and P. Lanusse et al. Mechatronics 57 (2019) 109–128

Fig. 22. Opening and closing valve at 6000rpm: refer-

ence signal (__ ); experimental response (__ ).

Fig. 23. Control voltage at 6000 rpm.

Fig. 24. Structure of the new control system

law.

122
Z. Dimitrova, M. Tari and P. Lanusse et al. Mechatronics 57 (2019) 109–128

Fig. 25. Pressure force (_ _ ); force generated by the actuator (__ ); reference position valve at 6000 rpm (__ ). (For interpretation of the references to color in the text,
the reader is referred to the web version of this article.)

Fig. 26. Opening and closing of the intake

valve at 6000rpm: reference signal (__ ); simu-
lation response (_ _ ).

Fig. 27. Control voltage at 6000 rpm (intake

simulation).

123
Z. Dimitrova, M. Tari and P. Lanusse et al.
4.3. Linear feedforward
Even if the feedback control system can ensure a tracking of the ref-
erence value zref by zmes that could be judged acceptable, taking into
account the admissible control effort and the fast variations of zref , the
response time of zmes is sometimes not as short as required. Thus, to
reduce this response time, a feedforward control system is used to pro-
vide a complementary control effort. A linear feedforward is based on
the inverse transfer function of the plant model (3):
( ( )2 )( ( )2 )
2𝜉 𝑠 2𝜉 𝑠
𝑠 1 + 𝜔 10 + 𝜔 𝑠 1 + 𝜔 20 + 𝜔 𝑠
𝑛10 𝑛10 𝑛20 𝑛20
𝐹 𝐹 (𝑠 ) = ( ) ∗ 𝐹 𝐶 (𝑠 ) (23)
𝐾0 1 + ∕𝜔𝑠
𝑝0
FC(s) is a low-pass filter used to make FF(s) proper and to provide
a control effort that satisfies the limitation on the plant control-effort
(48 V) for the highest possible variations of zref (related to 6000 rpm):
1 the end of the intake cycle. This strategy consists in applying a constant
𝐹𝐶 = ( )5 with 𝜔𝐹 𝐹 = 20000 rad∕s. (24)
1 + 𝑠∕𝜔
𝐹𝐹
4.4. Control law validation
Experimental validation results depicted in Figs. 20–23 of an open-
ing and closing valve at 1000 rpm and 6000 rpm confirm the effective-
ness of the developed control law in ensuring a relative good tracking
of the reference position with consistent performance for several valve
motion trajectories (repeatability). Nevertheless, the tracking obtained
at 6000 rpm could be judged not accurate enough by demanding users.
4.5. Nonlinear feedforward and plating strategy
To improve performance for high engine speeds and to reduce the
effect of the cylinder internal pressure on the output of the exhaust cam-
less valve actuator, since the controlled system has a notable nonlinear
behavior, a nonlinear feedforward control is necessary. The new feed-
forward strategy is based on the inversion of the dynamic model of the
actuator in order to obtain at each moment a balance of the forces hav-
ing an influence on the movement of the valve. The resultant Fact of
the forces acting on the actuator given by (2) (which also includes the
friction force) is used by the online calculation (in real time) of the cor-
responding current I using an order 5 polynomial equation that approx-
imates the electromagnetic model of the actuator depicted by Fig. 4.
No hysteresis phenomenon have been taken into account. A recorded
estimation of the pressure force Fpres_est is used into the Fact calculation.
Using a strictly-proper inverse model of the electrical part of the model,
it is easy then to deduce the necessary feedforward control voltage that
should ensure the valve position reference tracking and the rejection of
124
Mechatronics 57 (2019) 109–128
Fig. 28. Opening and closing of the exhaust
valve at 6000rpm: reference signal (__ ); simu-
lation response (_ _ ).
Table 1
Experimental seating valve velocity at various engine
speed.
Engine speed (rpm) Seating valve velocity (m/s)
Experimental Limitation (criteria)
1000 0.18 0.05
2000 0.3 0.1
3000 0.11 0.2
4000 0.17 0.3
5000 0.32 0.4
6000 0.41 0.5
the effect of the internal pressure in the cylinder. Furthermore, a plat-
ing strategy is added in order to maintain the valve closed especially at
and non-zero control signal outside the opening phases of the valve, out-
side these phases, the control signal remains that of the valve position
reference tracking. The new control law architecture of the system is
presented on Fig. 24.
Figs. 26–29 represent the simulation results for intake and exhaust
valve control at 6000 rpm.
On Fig. 24 are depicted the pressure force profile in the engine cylin-
der (Fpres (t)) and the electromagnetic force delivered by the actuator
(Fact (t)) for the exhaust valve at 6000 rpm. It should be noted that the
peak pressure force is observed after the exhaust valve closing and the
actuator is able to overcome the pressure force at the opening valve
(green line on Fig. 25).
Figs. 30–33 show the obtained experimental results valve at 1000
and 6000 rpm. These two tests are only significant of the intake phases
as the experimental testing bench is not able to apply a resistant force
close to the pressure forces observed for exhaust phases. We observe
that the robust feedback and nonlinear feedforward control law struc-
ture improves significantly the performance of the tracking valve posi-
tion, whereas the performance was poor at 6000 rpm with the robust
feedback and linear feedforward control law.
Finally, in order to validate the proposed control law structure, ex-
perimental valve seating velocities are shown on Table 1 for various
engine speed. It is demonstrated that, except for the low engine speed
(<2000 rpm) where the speed estimations are very noisy, the obtained
seating velocities fulfil the imposed limitations for a smooth valve seat-
ing, which ensure a “soft-touch” behavior. All these simulations and ex-
perimental tests mean that the designed controller ensure a good level
of precision to drive the actuators under realistic conditions. The pro-
posed method for the control law development is convenient for the
innovative actuator design.
Z. Dimitrova, M. Tari and P. Lanusse et al. Mechatronics 57 (2019) 109–128

Fig. 29. Control voltage at 6000 rpm (exhaust

simulation).

Fig. 30. Opening and closing of the intake

valve at 1000rpm: reference signal (__ ); experi-
mental response (__ ).

Fig. 31. Control voltage at 1000 rpm.

125
Z. Dimitrova, M. Tari and P. Lanusse et al.
Fig. 33. Control voltage at 6000 rpm.
5. Conclusion
An Lorentz force based electromagnetic actuator was developed and
experiments are used to identify unknown model parameters for robust
controller design. Simulations and experimental results show that a con-
trol law structure with a nonlinear feedforward and a plating strategy
allow a real improvement of the convergence valve trajectory to a de-
sired trajectory and make the actuator less sensitive to disturbances and
parameter variations. The article presents an interesting approach cou-
pling experimental measures and a simulation design. The development
of the control law is based on an experimentally validated control model.
The precision of the actuators and the robustness of the control law
are important parameters for the integration of the actuators in the en-
gine cylinder head. The electromagnetic actuators give the possibility
to manage the engine gas exchange on a very performant way. The
precision of the control law attests the controllability of the actuators
126
Mechatronics 57 (2019) 109–128
Fig. 32. Opening and closing of the intake
valve at 6000rpm: reference signal (__ ); experi-
mental response (__ ).
technology and thus the good performances of the engine. The robust
and precise controlling of the actuators on the intake side is especially
important for the regulation of the airflow in the engine and this for
gasoline engine concerns directly the regulation of the engine torque.
The respect of the velocity requirements during the valve landing is an
important step and insures the respect of the noise criteria and the me-
chanical durability of the system.
These results are promising for the development of the engine con-
trol law that manages the gas exchanges of the engine. Future steps of
the work could be the integration of the actuators in the engine cylin-
der head, the integration of the electronics and the development of the
control strategies for full Camless engine.
Appendix A
Fig. A3.
Z. Dimitrova, M. Tari and P. Lanusse et al.
Fig. A1. Schematic of the actuator, with magnetic circuit.
Fig. A2. Actuator in opened position.
127
Mechatronics 57 (2019) 109–128
Fig. A3. Actuator in closed position.
Bibliography
[1] Liu L, Chang S. Improvement of valve seating performance of engine’s electromag-
netic valvetrain. Mechatronics 2011;21(7):1234–8.
[2] Mercorelli P. A Lyapunov-based adaptive control law for an electromagnetic actua-
tor. AASRI Procedia 2013;4:96–103.
[3] Hoffmann W, Stefanopoulou A. Valve position tracking for soft landing of electrome-
chanical camless valvetrain. IFAC Proc 2001;34(1):295–300.
[4] Peterson K, Stefanopoulou A, Wang Y, Haghgooie M. Nonlinear self-tuning
control for soft landing of an electromechanical valve actuator. IFAC Proce
2002;35(2):191–6.
[5] Liu J, Pien Yang Y, Hong Xu J. Electromechanical valve actuator with hybrid MMF
for camless engine. IFAC Proc 2008;41(2):10698–703.
[6] Bolat FC, Sivrioglu S. Active control of a small-scale wind turbine blade containing
magnetorheological fluid. Micromachines 2018;9(2):80.
[7] Bolat FC, Sivrioglu S. Active vibration suppression of a flexible blade element using
magnetorheological layer patch-electromagnetic actuator. Turkish J Electro Mech
Energy 2018;3(June(1)).
[8] Tai C, Tsao TC, Levin MB. Adaptive nonlinear feedforward control of an electrohy-
draulic camless valvetrain. In: Proc. Amer. Control Conf., 1–6; 2000. p. 1001–5.
[9] Pournazeri M, Khajepour A, Huang Y. Improving energy efficiency and robustness of
a novel variable valve actuation system for engines. Mechatronics 2018;50:121–33.
ISSN 0957-4158. https://doi.org/10.1016/j.mechatronics.2018.02.002.
[10] Werner N, Mercorelli P, Becker U, Harndorf H, Niekerk TV. Digital control of a
camless engine using Lyapunov approach with backward Euler approximation. IFAC
Proc 2014;47(3):5679–84. https://doi.org/10.3182/20140824-6-ZA-1003.02562.
[11] Duchaug JL. Design, modeling and dimensionning of an electromagnetic variable
valve lift 3-stage tappet in an automotive engine Ph.D. Thesis. France: ENS Cachan;
2015. https://www.researchgate.net/publication/281471347.
[12] Zammit JP, McGhee MJ, Shayler PJ, Pegg I. Benefits of cylinder deactivation on
a diesel engine and restrictions due to low boost. In: Internal combustion engines:
performance, fuel economy and emissions. IMechE, Woodhead Publishing; 2013.
p. 95–108.
[13] Dimitrova Z, Maréchal F. Gasoline hybrid pneumatic engine for efficient vehicle
powertrain hybridization. Appl Energy 2015;151:168–77.
[14] Zhao J, Seethaler RJ. A fully flexible valve actuation system for internal
combustion engines. IEEE/ASME Trans Mechatronics April 2011;16(2):361–70.
doi:10.1109/TMECH.2010.2043850.
[15] Wong PK, Tam LM, Li K. Modeling and simulation of a dual-mode electrohy-
draulic fully variable valve train for four-stroke engine. Int J Autom Technol
2008;9(5):509–21.
[16] Eyabi P, Washington G. Modeling and sensorless control of an electromagnetic valve
actuator. J Mechatronics 2006;16:159–75.
[17] Cope D, Wright A. Electromagnetic fully flexible valve actuator; 2006. SAE Technical
Paper 2006-01-0044 https://doi.org/10.4271/2006-01-0044.
[18] Ren H, Wang T. Development and modeling of an electromagnetic energy harvester
from pressure fluctuations. Mechatronics 2018;49:36–45.
Z. Dimitrova, M. Tari and P. Lanusse et al.
[19] Dimitrova Z, Porto Mendes LG, Kossowski Z. Electromagnetic actuator for a valve
of an internal combustion engine. PSA Peugeot Citroen; 2016. patent number
WO2016051041.
[20] Tai C, Tsao TC. Control of an electromechanical actuator for camless engine. In:
Proc. Amer. Control Conf., 16; 2003. p. 3113–18.
[21] Haskara I, Kokotovic V, Mianzo L. Control of an elctromechanical valve actuator for
a camless engine. Int J Robust Nonlinear Control 2004;14:561–79.
[22] Peterson KS, Grizzle JW, Stefanopoulou AG. Nonlinear control for magnetic levita-
tion of automotive engine vales. IEEE Trans Conrols Syst Technol 2006;14(Febru-
ary(2)):346–54.
[23] Montanari M, Ronchi F, Rossi C, Tonielli A. Control of a camless engine electrome-
chanical actuator: position reconstruction and dynamic performance analysis. IEEE
Trans Ind Electron April 2004;51(2):299–311. doi:10.1109/TIE.2004.825230.
[24] Peterson KS, Stefanopoulou AG. Extremum seeking control for soft landing of an
electromechanical valve actuator. Automatica 2004;40:1063–9.
[25] Oustaloup A. Systèmes asservis linéaires d’ordre fractionnaire. Paris: Masson; 1983.
[26] Oustaloup A. La commande CRONE. Paris: Hermes Editor; 1991.
[27] Oustaloup A, Mathieu B, Lanusse P. The CRONE control of resonant plants: Appli-
cation to a flexible transmission. Eur J Control 1995;1(n° 2):113 121.
[28] Lanusse P. De la commande CRONE de première génération à la commande CRONE
de troisième generation Ph.D. Thesis. France: Bordeaux I University; 1994.
[29] Lanusse P. CRONE control system design, a CRONE toolbox for Matlab; 2010
http://www.ims-bordeaux.fr/CRONE/toolbox.
[30] Seethaler R, Duerr K. Electromagnetic valve actuation with two configurations. In:
Proc. ASME IDETC/CIE; 2009.
128
View publication stats
Mechatronics 57 (2019) 109–128
[31] Sabatier J, Lanusse P, Melchior P, Oustaloup A. Fractional order differentiation and
robust control design. Springer Academic edition; 2015.
[32] Landau ID, Rey D, Karimi A, Voda A, Franco A. A flexible transmission system as a
benchmark for digital control. Eur J Control 1995;1(n° 2).
Dr. Dimitrova is an Innovation Project Leader in the R&D
Division of Groupe. PSA Prior to assuming this position she
had been Innovation Project Manager at the Scientific and Re-
search Department of PSA Peugeot Citroën, since June 2012.
She joined PSA Peugeot Citroën in May 2006 from Porsche
AG. Zlatina Dimitrova served in various R&D functions at PSA
Peugeot Citroën for 12 years. She has various technical and
research experiences in the R&D Division of PSA, where she
started as combustion engineer.
Dimitrova, holds a Ph.D. degree in mechanical engineering
from the Ecole Polytechnique Fédérale de Lausanne, Switzer-
land. She holds as well an engineering degree from the Insti-
tut National des Sciences Appliquées (INSA) in Lyon, France,
with a degree of mechanical engineering. She is also graduated
Master in Sciences in Mechanical engineering with specializa-
tion on combustion engines and turbomachines from the Karl-
sruhe Institute of Technology (KIT), Germany.