Content uploaded by Valentin Ivanov
Author content
All content in this area was uploaded by Valentin Ivanov on Jun 16, 2016
Content may be subject to copyright.
Page 1
EB2016-SVM-076
ADVANCED BRAKING SYSTEM CONTROL PROTOTYPING USING
NETWORKED HARDWARE-IN-THE-LOOP TECHNIQUE
1Schreiber, Viktor; 1Savitski, Dzmitry; 1Augsburg, Klaus; 1Ivanov, Valentin; 2Schyr, Christian
1Technische Universität Ilmenau, Germany; 2AVL Deutschland GmbH, Germany
KEYWORDS – braking system control, x-in-the-loop, hardware-in-the-loop, software-in-
the-loop, model-in-the-loop, real-time, test rig
ABSTRACT – Control functions for the base-braking and emergency braking situations are
important element of the vehicle active safety and have high requirements to robustness. The
corresponding control algorithms should be reliable, provide sufficient level of system
adaptiveness and be able to reject external disturbances. This demands not only the well-
organized controller from the theoretical point of view, but also its systematic experimental
validation. Moreover, effects and factors, which can potentially produce deterioration of
braking system control functions, should be properly taken into account in the simulation and
during the experiments. Another important factor is that brake control systems have a closed-
loop operation with the tyre-road interaction, and its operation is accompanied by such
complex effects like (i) variation of disc/pad friction coefficient and (ii) brake hysteresis. This
produces strong demand on extension of the conventional testing facilities for the braking
system control evaluation. Therefore, besides the part of the control system design, this paper
represents possible advancement of hardware-in-the-loop testing procedure for development
and validation of braking system control functions.
INTRODUCTION
Current tendencies of braking systems development are moving towards x-by-wire
architecture and enhancement in actuator performance. Such solutions allow simultaneous
enhancement in vehicle safety, comfort and energy efficiency by avoiding traditional trade-
offs between these characteristics. Due to these reasons, the brake-by-wire architecture has a
demand on new arts of control systems design and requires more sophisticated control
algorithms.
Potential of advanced braking system control has been recently investigated in a number of
publications. In particular, the paper [13] has addressed the minimization of pitch oscillations
during the braking. The corresponding approach has been proposed for a decoupled electro-
hydraulic brake system by offline optimization of the braking forces distribution that allowed
achieving better ride comfort. Another system variant related to the improvement of
longitudinal ride comfort has been investigated in paper [15] for the control algorithm aimed
at reduction of longitudinal vehicle jerk during service braking. Next example considers
simultaneous operation of different braking actuators (e.g. electric motors and hydraulic
brakes) to receive benefits in terms of energy efficiency and braking performance. This
example is discussed in paper [16], where the reduction of braking distance on 20% was
achieved by application of continuous anti-lock braking control strategy and proper blending
of the braking torque between electric motors and hydraulic brakes. Another kind of
functional advancement is introduced in [14]. This study has proposed to attenuate judder
oscillations in electro-mechanical brakes using an adaptive control approach. A number of
Page 2
studies are also investigating the functionality of the braking systems in the control loop
beyond a stand-alone operation. For instance, the paper [19] shows the integration of several
active chassis systems in framework of integrated chassis control, which resolves the known
issue with control effects overlap of over-actuated systems.
Such shift in principles of braking system control design must be supplemented with
appropriate modifications of prototyping procedure. Next section of this paper will introduce
investigations, where influence of several physical phenomena can produce deterioration of
braking performance. By means of brake-by-wire systems they can be attenuated, but as
close-loop control system it requires using of more complex testing procedure, demand in
which will be shown on some examples. After this section proposed prototyping technique
will described in details.
ADVANCED BRAKING SYSTEM CONTROL
To fully utilize the potential of x-by-wire systems, the control algorithm should consider
several physical phenomena, which can produce negative influence on the braking
performance. In recent years Automotive Engineering Group from Technische Universität
Ilmenau performed several investigations aimed at understanding of physical processes,
which influence the performance of braking control systems. In this section some of results
will be discussed and demand in further development of testing techniques will be shown.
Authors in [18] indicated that variation of the coefficient of friction in brake pads and disc can
distort the brake pedal feel characteristics. In the same paper the algorithm on attenuation of
such sort of disturbances was proposed through estimation of vehicle state and parameters and
establishing the correction factor for compensation of forces generated by variation of brakes
coefficient of friction, road grade and vehicle mass. The testing procedure has included (i)
separated investigation of the coefficient of friction variation on the dynamometric test rig,
and (ii) validation of the vehicle state and parameter observer on the real vehicle. Further, all
these outcomes were used to test the disturbance attenuation algorithms on the hardware-in-
the-loop test rig with installed decoupled braking system. As it can be seen, validation of such
closed-loop algorithms demands a multi-domain real-time testing procedure.
In respect to the judder oscillations of brakes and their compensation, the paper [14] has
investigated possibilities of the electro-mechanical and electro-hydraulic actuators to attenuate
low-frequency disturbances caused by disc thickness variation (DTV). At the test rig with
decoupled EHB system reference information about DTV data has been utilized and allowed
to prove applicability of developed control algorithms.
In the paper [3] investigation of brakes behaviour and its influence on anti-lock braking
system has been discussed. It offers combined testing technique, where two test rigs, namely
hardware-in-the-loop and dynamometric test rig, were physically coupled. It evidently
showed influence of the physical processes in friction brakes on the control performance
producing phase shifts in ABS. This research was aimed not at investigation of physical
phenomenon, but at showing demand in coupling/networking of test rigs.
Influence of tire characteristics on ABS performance was studied in [17]. It shows
considerable changes in the braking performance due to variation of tire pressure. In absence
of reliable tire models for simulation, prototyping and adjustment of the ABS could be done
before the road tests if braking system and tire testing rig can be coupled together.
Outcomes of aforementioned papers in relation to the concept of networked testing
environment were summarized in Figure 1. In these studies, authors were forced to perform
Page 3
testing procedures separately from each other or
move straightforward to the road tests. It applies
several limitations on testing of control algorithms,
which use reference experimental results instead of
direct information receipt from vehicle subsystems.
This gap can be fulfilled by networking the testing
environment for validating closed-loop control
algorithms and investigation of physical
phenomenon influence on brake control systems.
In the next section of paper, it will be discussed
how to break some certain limitations in
development, testing and validation procedures
though new experimental approach.
PROTOTYPING TECHNIQUE
The proposed methodology for developing of braking systems is based on coupled test rigs,
embedded hardware, and real-time simulation tools. This technique was already investigated
partially in some previous publications, for instance, in [3, 7, 8, 10, 11, 12]. However, until
today there are no common philosophies or standards to describe this methodology. In order
to clarify and to ensure a consistent definition of terms, the concept is explained and its
motivation is given in more detail.
Motivation – Despite the increasing mechanization, mechatronization and complexity of
subsystems in ground vehicles, the technical benefits converge rapidly. The development
process of vehicle components and other technical products are still designed in stand-alone
operation mode. By coupling relevant vehicle subsystems in numerous physical, computing,
electrical and relevant cyber-physical domains the benefits for the system design will be
significantly multiplied. Consequently, e.g. on the field of active safety and driver assistance
systems, a higher effect can be achieved if all the subsystems are coordinated. For such
systems there is a strong requirement for novel testing techniques and developing
environments to improve their functionality.
Besides that, it is expected that this methodology will enable a higher level of basic research
on multidisciplinary subject areas under reproducible conditions. Accordingly complex
relationships (e.g. between brake friction behaviour and vehicle dynamics control) can be
investigated. For these reasons, new methods, procedures and standards are desirable to drive
future advanced system for ground vehicles preceded.
Clarification of concept – As it was already mentioned, the anticipated prototyping
technique is established on networked test rigs across different domains. Beyond that, more
tools and equipment are needed to perform advanced functionality, simulate virtual vehicular
systems or evaluate software code. Commonly these developing and testing platforms are
considered into the three main categories: hardware-in-the-loop, software-in-the-loop and
model-in-the-loop [9]. In [12] it was proposed by authors to extend these elementary
categories to a test-rig-in-the-loop concept to close the gap between embedded computing
systems and physical test beds respectively. Thereby, the “In-The-Loop” convention refers to
the feedback of the response back to the input. The major terms are defined as follows:
Model-In-The-Loop (MiL) – On the very early stage of designing cyber-physical systems
Figure 1
: Factors influencing on braking
system control performance
Page 4
simulation models are used for verifying basic functionality and behavior. These simulation
models are executed on higher leveled software systems which are abstractly related to the
actual system.
Software-In-The-Loop (SiL) – For software evaluation, a compiler generates a software code
out of the simulation model or parts of it. The generated code is running instead of the higher
leveled simulation software or their parts.
Hardware-In-The-Loop (HiL) – To execute the developed functionality under realistic
conditions like peripheral equipment or stimulus signals, the compiled software code is
loaded and running on a real-time embedded system. The HiL is regularly used for measuring,
(signal) processing, communicating or controlling tasks. HiL is also called processor-in-the-
loop (PiL).
Test-Rig-In-The-Loop (TRiL) – Finally, at the end of the developing chain, a Test-Rig-In-The-
Loop is needed to verify and proving the functionality. TRiL represents the physical part of
the developed systems, e.g. a brake system.
Table 1: Overview of closed-loop prototyping techniques
In-The-Loop
Category
Model-In-The-
Loop
Software-In-The-
Loop
Hardware-In-
The-Loop
Test-Rig-In-The-
Loop
Real-time
capability
No/partially
No/partially yes Yes
Developing stage very early early proceeded late evaluation
examples Matlab/Simulink
IPG CarMaker
Microtec PowerPC C
Compile
GNU C Compiler
dSpace RT-
hardware
AVL InMotion
Brake
Dynamometer
Vehicle Chassis
Dynamometer
application vehicle
simulation model
Emulation of
protocols/functionality
CAN I/O, Voltage
I/O, signal
processing
ABS, ESP, brake
dust investigations,
NVH
X-In-The-Loop – Superordinate all featured In-The-Loop techniques are summarized in X-In-
The-Loop. With this background XiL is a real-time holistic networking requiring
interdisciplinary approach (Physics, IT, Control Engineering, Systems Analysis, etc.). In this
form all XiL participants are modular combinable to optimize the outcomes of testing
procedures. Therefore, a generic real-time interface is required to exchange configurations,
parameters, inputs and outputs between applicants. A substantial flexibility is expected from
this approach because all participants come/derived from different developing stages. This
allows a rapid prototyping through all developing stages.
Figure 2: Overview of the superordinate XiL configuration at the Technische Universität Ilmenau
Use case configurations for prototyping advanced braking systems
Currently Technische Universität Ilmenau is implementing the proposed prototyping
technique. As a very first step, a Vehicle Chassis Dynamometer (AVL), Brake Dynamometer
Page 5
(Link Engineering), Embedded System (dSpace) and Real-Time Simulation (InMotion/IPG-
CarMaker) should be connected. Initially, studies show a significant benefit in application of
this testing technique.
Nevertheless, there are also still challenges in case of various number of degrees of freedom.
In respect to study about braking systems prototyping, only one brake dynamometer was used,
i.e. single DOF. But, for complex investigation like ABS performance for full vehicle
dynamic, one DOF for each wheel is essential. Hence, the single reference is overdetermined.
This become challenging due to an exchange between coupled real-time systems of different
domains. For this issue a method for transformation of relevant system state from single to
multi-degree of freedom was investigated. Figure 3 shows the basic conception, which will be
explained using the example of brake systems. As illustrated, all wheel speeds are used to
transform them to only one reference. This single reference is used to control the
dynamometer. For controlling the brake pressure the pressure of the master cylinder is used.
The responding braking torque and the temperature are reverse transformed and feedback to
the RT-Simulation.
EXPERIMENTAL RESULTS
The performed experiments aimed at showing demand in coupling of the testing facilities for
development of active control functions for brake systems. In this experiment following
testing facilities were used: (i) AVL InMotion platform with IPG CarMaker software, (ii)
dynamometric test rig and (iii) hardware-in-the-loop test rig equipped with electro-hydraulic
brake system. The principal scheme of test rigs communication operating as closed-loop
system is shown in Figure 3.
Figure 3: Brake Dynamometer coupled with an Embedded RT-Simulation
Page 6
Considering known communication technologies and industrial communication standards for
networking controller units, sensors and actuators, it is suggested to connect real-time
platforms with Ethernet protocol. The Ethernet standard covers a high bandwidth and low
communication latency, as it is represented in Figure 4. Due to this reason the proposed XiL
environment was connected via User Data Protocol (UDP).
Figure 4: Performance of common communication standards
To investigate the performance of the communication network, aforementioned test rigs from
different physical domains were crosslinked and the most remote of them (AVL InMotion
platform and dynamometrical test rig) are located in different buildings with a distance of
approximately 1 km. Real-time platform and control PCs at each end have an appropriate
networking device, which supports Ethernet communication. According to the use case
description in Figure 2 the Real-Time-Master (AVL InMotion) distributes reference inputs,
namely braking pressure demand and wheel speed, to other participants (Slave) in the XiL
Setup. The Slaves respond with an input for the Master-Device (in our case actual braking
torque). In terms of communication quality, it was figured out, that the buffer of the
networking devices accumulates rapidly after a few seconds, which leads to a steadily
increasing communication delay, Figure 4. To solve this issue, it was decided to include
pauses for handling the buffer processing as it is shown in Figure 5 (b). The processing of the
network device buffer provides reliable real-time capability for frequencies up to 250 Hz.
With such communication performance an adverse effect on operation of networked testing
facilities is not expected. Therefore, this networking technique offers an opportunity in design
and investigation of multi-domain physical and mechatronic systems, as the complete brake
system (hydraulic lines, control, friction brakes etc.) in our case.
Figure 5: Communication delay of networking devices
100101102103104
10-1
100
101
102
103
bandwidth [
MBit
/
s
]
latency [µs ]
Fast Ethernet
Gigabit-Ethernet
USB 2.0
IEEE 1394a
CAN Highspeed
P CI Express (x4)
P CI/PXI (x4)
M OST
0 20 40 60 80 100 120
1
2
3
4
5x 10
-3
t [s]
communication delay [ s]
(b) networking device achieves to process buffer
0 20 40 60 80 100 120
0
0.005
0.01
0.015
t [s]
communication delay [s ]
(a) networking device is not able to process the full puffer
Page 7
The dynamometric test rig is used in this experiment to deliver information about the wheel
torque to the simulation software at AVL InMotion platform. Compared to the simulation
solutions, it considers tribological phenomena in brakes. As it known, the friction behaviour is
depending on many factors such as the temperature of the friction area, relative velocity
between friction components, pressure level (pressure distribution in the contact area) and
pressure gradient.
Besides these factors, the brake system is characterized by hysteretic losses in friction pair,
which cannot be neglected during design of brake control system. The commonly used
mathematical models for modelling of hysteresis, like Preisach [20, 21], Jiles-Atherton [22] or
Bouc-Wen [23, 24, 25], were initially designed for ferromagnetic problems and as any semi-
empirical model does not consider all physical effects, especially if speak about brake system.
Such approach in its initial form is not applicable for claims as a robust controller design.
As it is shown in Figure 7, the friction coefficient increases with rising temperature in the
contact zone. However, it cannot be seen that the friction coefficient rises up to a maximum
value. After this critical point the friction coefficient collapses until the so-called fading. In
this case the organic ingredients of the friction linings become gaseous. Beyond, the friction
coefficient indicates significant hysteresis behaviour while a dynamic actuating. In Figure 6 it
is shown, that the hysteresis loop is getting wider with increasing actuating speed of the brake
pressure. It is assumed that this behaviour is basically caused by tribological phenomena.
Figure 6: Impact of brake pressure gradient on
Hysteresis characteristics of friction brakes
(temperature, speed = const.)
Figure 7: Hysteresis characteristics of friction brakes in
dependence on temperature, brake pressure level and
brake pressure gradient (speed = const.)
As it was mentioned in introduction, the ABS control is evidently dependent on the friction
behaviour of brakes. The electro-hydraulic brake system at the testing platform uses the
signals about brake pressure demand and wheel speeds realising ABS control functions. In the
investigated cases it was found that hysteresis has a vital influence on braking performance
during straight-line ABS braking especially on low-µ surfaces.
The velocity profile in Figure 8 clearly shows that braking time was increased on 2 seconds
compared to the case, where hysteresis effects were not considered. It is caused by delays in
brake torque generation and phase shifts in ABS control, as it is shown in Figure 9.
Such braking distance variation and deterioration of ABS function caused by influence of
physical phenomena in brakes shows a strong demand in novel testing procedures. Proposed
remote connection of test rigs provides new opportunities in combined testing techniques,
where test rigs are located even on relatively long distance between each other.
10 15 20 25 30 35 40
0.23
0.24
0.25
0.26
0.27
0.28
0.29
0.3
0.31
0.32
[bar]
µ
[-]
1Hz
2Hz
10 15 20 25 30 35 40
0.25
0.26
0.27
0.28
0.29
0.3
0.31
0.32
0.33
0.34
0.35
[bar]
µ
[-]
ϑ
[
°
C]
40
50
60
70
80
90
100
110
120
Page 8
Figure 8: Speed profile for ABS braking without (blue
line) and with (red line) consideration of brake hysteresis
Figure 9: Brake pressure profile for ABS braking
without (blue line) and with (red line) consideration of
brake hysteresis
IMPACT AND CONCLUSION
Benefit – Obtained results in both areas of braking system control design and its testing by
means of coupled test rigs show excellent opportunity to shorten the step between simulation
and road tests with simultaneous enhancement of system performance, adaptability and
robustness. Besides that, the research and investigation of complex coherent issues is under
reproducible and steady conditions. By linking specialized testing, measuring, controlling and
simulation systems will increase the precision of outcome of domain specified functionality,
e.g. connected brake dynamometer with an embedded simulation and control system.
Future Challenges – Caused by growing quantity of prototyping platforms in the XiL-
Environment and different locations the effort for preparation and configuration is increased.
Moreover, the prototyping platforms such as test rigs (TRiL), embedded software (HiL/SiL)
and advanced simulation (MiL) from different vendors are usually not applicable. An
implementation of a centralized allocation system for distributing the configuration data will
simplify the handling of information exchange and performing test procedures.
ACKNOWLEDGE
The proposed prototyping technique is represented on its initial stage of development and will
be further evolved and tested in the following projects:
ACOSAR – The European founded project purpose the development a non-proprietary
“Advanced Co-simulation Interface” (ACI) for RT-System integration and an according
integration methodology, which shall be a substantial contribution to international
standardization (FMI).
EVE – This work is supported by the European Union Horizon 2020 Framework Program,
Marie Skłodowska-Curie actions, under grant agreement no. 645736. The research objectives
are focused on the development of experimental tyre database that can be used in the design
of new chassis control systems, advanced models of ground vehicles and automotive
subsystems for real-time applications, and novel integrated braking control methods.
VISTA4F – This project is founded by the Thuringian Ministry TMBWK. Within the project,
the use-case scenarios will be demonstrated.
Page 9
REFERENCES
1. Savitski, D., Hoepping, K., Ivanov, V., and Augsburg, K., Influence of the Tire Inflation
Pressure Variation on Braking Efficiency and Driving Comfort of Full Electric Vehicle
with Continuous Anti-Lock Braking System, SAE Int. J. Passeng. Cars - Mech. Syst.
8(2):460-467, 2015.
2. Ivanov, V., Savitski, D., Augsburg, K., Knauder, B., Barber, P., Zehetner, J. Wheel slip
control for AWD electric vehicle, 18th International Conference of the International
Society for Terrain-Vehicle Systems, Seoul, Korea, 2014 (Best Paper Award).
3. Augsburg, K., Savitski, D., Heidrich, L., and Ivanov, V., Combined Testing Technique:
Development of Friction Brake System for Electric Vehicle. SAE Technical Paper 2014-
01-2529, 2014, doi:10.4271/2014-01-2529.
4. Ivanov, V., Savitski, D., Augsburg, K., Barber, P. Electric Vehicles with Individually
Controlled Motors: Revisiting the ABS Design, IEEE International Conference on
Mechatronics, 323-328, Nagoya, Japan, 06-08.03.2015 (Best Paper Award).
5. Savitski, D., Ivanov, V., Augsburg, K., Shyrokau, B., Wragge-Morley, R., Pütz, T.,
Barber, P. The New Paradigm of Anti-Lock Braking System for Full Electric Vehicle:
Experimental Investigation and Benchmarking, Proceedings of the Institution of
Mechanical Engineers: Part D-Journal of Automobile Engineering, 2015
6. Savitski, D., Ivanov, V., Heidrich, L., Augsburg, K., Pütz, T. Experimental investigation
of braking dynamics of electric vehicle, Eurobrake 2013 Conference, 17-19.06.2013,
Dresden, Germany.
7. De Souza, I.D.T.; Silva, S.N.; Teles, R.M.; Fernandes, M.A.C. Platform for Real-Time
Simulation of Dynamic Systems and Hardware-in-the-Loop for Control Algorithms.
Sensors 2014, 14, 19176-19199. http://dx.doi.org/10.3390/s141019176
8. Albers, Albert; Behrendt, Matthias; Schroeter, Jens; Ott, Sascha; Klingler, Simon. X-in-
the-loop: A framework for supporting central engineering activities and contracting
complexity in product engineering processes, 19th International Conference on
Engineering Design (ICED13), 19-22.08.2013, Seoul, Korea.
9. VDI 2206, VDI, Design methodology for mechatronic systems, June 2004.
10. Albers, A.; Dueser, T.; Sander, O.; Roth C., Henning, J.: X-in-the-Loop-Framework für
Fahrzeuge, Steuergeräte und Kommunikationssysteme, ATZelektronik, Vol. 5(5), pp. 60-
65, 2010.
11. Schröter, J.: The advanced X-in-the-Loop-Framework for continuous integration of
optimization procedures into the product development process using examples of the
development of fuel efficient vehicles, PhD thesis, IPEK, Karlsruhe, Germany, 2013.
12. Augsburg, K., Gramstat, S., Horn, R., Ivanov, V. et al., "Investigation of Brake Control
Using Test-Rig-in-the-Loop Technique," SAE Technical Paper 2011-01-2372, 2011,
doi:10.4271/2011-01-2372.
13. Savitski, D., Augsburg, K., Ivanov, V. Enhancement Of Energy Efficiency, Vehicle
Safety And Ride Comfort For All-Wheel Drive Full Electric Vehicles, EuroBrake 2014,
Lille, France, 2014.
14. Lee, C. F. and Manzie C., "Adaptive Brake Torque Variation Compensation for an
Electromechanical Brake," SAE Journal of Passenger Cars - Electronic and Electrical
Systems, 5(2):600-606, SAE Technical Paper 2012-01-1840, 2012.
Page 10
15. Barber, P. and Williams H., "Chauffeur braking," ESGI68, Leeds, United Kingdom, 2009.
16. Savitski, D., Ivanov, V., Augsburg, K., Shyrokau, B., Wragge-Morley, R., Pütz, T.,
Barber, P. The new paradigm of an anti-lock braking system for a full electric vehicle:
experimental investigation and benchmarking, Proceedings of the Institution of
Mechanical Engineers, Part D: Journal of Automobile Engineering, 2015,
doi:10.1177/0954407015608548.
17. Savitski, D., Hoepping, K., Ivanov, V., Augsburg, K. Influence of the Tire Inflation
Pressure Variation on Braking Efficiency and Driving Comfort of Full Electric Vehicle
with Continuous Anti-Lock Braking System, SAE Int. J. Passeng. Cars - Mech. Syst.
8(2):2015, doi:10.4271/2015-01-0643
18. Savitski, D., Shyrokau, B., Ivanov, V. Base-Brake Functions of Electric Vehicle:
Disturbance Compensation in Decoupled Brake System, International Journal of Vehicle
Design, 2015.
19. Shyrokau, B., Wang, D., Savitski, D., Hoepping, K., Ivanov, V. Vehicle Motion Control
with Subsystem Prioritization, International Journal of Mechatronics, Vol.30 pp. 297–315,
2015, doi:10.1016/j.mechatronics.2014.11.004
20. F. Preisach: Über die magnetische Nachwirkung. Zeitschrift für Physik, 94 pp. 277-302,
1935
21. Ruderman, M., Bertram, T.: Discrete Dynamic Preisach Model for Robust Inverse Control
of Hysteresis Systems, 49th IEEE Conference on Decision and Control, USA, 2010
22. Jiles, D. C., Atherton, D. L.: Theory of ferromagnetic hysteresis. Journal of Applied
Physics 55, 1984, doi: 10.1063/1.333582.
23. Bouc, R.: Forced vibration of mechanical systems with hysteresis, Proceedings of the
Fourth Conference on Nonlinear Oscillation, p. 315, Prague, 1967
24. Wen, Y. K.: Method for random vibration of hysteretic systems, Journal of Engineering
Mechanics (American Society of Civil Engineers) 102 (2), pp. 249-263, USA, 1976
25. Song, J., Der Kiureghian, A.: Generalized Bouc–Wen model for highly asymmetric
hysteresis, Journal of Engineering Mechanics. ASCE. Vol 132, No. 6, pp. 610-618, 2006