Aalborg Universitet

Mechatronic System Design Based On An Optimisation Approach

Andersen, Torben Ole; Pedersen, Henrik Clemmensen; Hansen, Michael Rygaard

Publication date:
2007

Document Version
Publisher's PDF, also known as Version of record

Link to publication from Aalborg University

Citation for published version (APA):

Andersen, T. O., Pedersen, H. C., & Hansen, M. R. (2007). Mechatronic System Design Based On An
Optimisation Approach. Paper presented at International Workshop - Mechatronics Day 2007, København,
Denmark.

General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners
and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

? Users may download and print one copy of any publication from the public portal for the purpose of private study or research.
? You may not further distribute the material or use it for any profit-making activity or commercial gain
? You may freely distribute the URL identifying the publication in the public portal ?
Take down policy
If you believe that this document breaches copyright please contact us at [email protected] providing details, and we will remove access to
the work immediately and investigate your claim.

Downloaded from vbn.aau.dk on: April 01, 2021

 MECHATRONICS DAY 2007 - International Workshop

MECHATRONIC SYSTEM DESIGN BASED ON AN OPTIMISATION

APPROACH

Torben O. Andersen Henrik C. Pedersen

Aalborg University Aalborg University
Institute of Energy Technology Institute of Energy Technology
DK-9210 Aalborg East, Denmark DK-9210 Aalborg East, Denmark
Phone +45 9635 9269, Fax +45 9815 1411 Phone +45 9635 9275, Fax +45 9815 1411
E-mail: [email protected] E-mail: [email protected]

Michael R. Hansen
Aalborg University
Institute of Mechanical Engineering
DK-9210 Aalborg East, Denmark
Phone +45 9635 9321, Fax +45 9815 1411
E-mail: [email protected]

ABSTRACT
The envisaged objective of this paper project is to extend the current state of the art regarding
the design of complex mechatronic systems utilizing an optimisation approach.
We propose to investigate a novel framework for mechatronic system design. The novelty and
originality being the use of optimisation techniques. The methods used to optimise/design
within the classical disciplines will be identified and extended to mechatronic system design.

KEYWORDS: Mechatronic System Design, Optimisation, Modelling, Simulation, Control.

1 INTRODUCTION
Throughout the last half century, the requirements for new products and systems have radically
changed, driven by the need of manufacturing more advanced, better and efficient products, in
order to stay competitive. Today it is no longer enough to design a product which fulfils its
intended task satisfactorily, it needs to be more efficient, versatile, unique and cost-effective
than other similar products in order to be competitive.
Mechatronic systems and products here pose new opportunities, as the combination of the
different technologies may not only open for designing simpler systems, where e.g. a complex
mechanical system is replaced by an actuator solutions, but may also bring with it new features
and functions, that could not earlier be realised with only one technology. The latter exempli-
fied through e.g. fault monitoring systems and active damping. Designing mechatronic systems
however pose new problems due to the interaction of technologies. There are many designs
where electronics and control are combined with mechanical components, but with very little

1
synergy and poor integration they just become a marginally useful, error-prone, expensive con-
glomeration. Synergism and integration in the design process is what set apart a Mechatronic
System from a traditional, multidisciplinary system, but this also adds new complexity to the
design process. The design process is in this way not only further complicated by the addition
and mixing of technologies, but also the by the circumstances that changing one or more design
parameters may highly influence a complete design, where the effects may not be foreseen up
front, but is left to be analysed afterwards, as the correlations are typically highly non-linear.
The typical design approach has been to divide the design problem into sub problems for
each technology area (mechanics, electronics and control) and describe the interface between
the technologies, whereas the lack of well-established, systematic engineering methods to form
the basic set-off in analysis and design of complete mechatronic systems has been obvious. This
lack of methods is among other complicated by the following reasons:

• Insight (theoretical) into each of the technical disciplines are required to fully understand
the possibilities and limitations in the different technologies and the advances and limita-
tions that may arise when interacting.

• When considering actuated mechatronic systems, i.e. systems that include moving parts,
many problems are of extremely dynamic nature and can only be described by differential
equations.

• Many components have inherently non-linear characteristics, making the describing dif-
ferential equations non-linear and thus limits the applicability of linear control theory.

• Mechanical systems and some actuation system types (hydraulic and pneumatic) are in
fact distributed parameter systems. It takes some skill to make sound engineering judge-
ments on when and how lumped parameter approaches can be made and/or when rigidity
(i.e. structural flexibility) significantly influence the dynamic behaviour.

• Finally the development of valuable engineering theories very often must be based on
experimental verification of hypothesis.

The problem or engineering effort in mechatronics is therefore both of educational and

technical nature.
It seems from the above statements, that strong efforts in the elevation of the engineering
methods ought to be emphasised as one mean to ensure that the full potential in mechatronic sys-
tems are utilised, hereby ensuring the companies a continuous ability to compete. Optimisation
methods combined with modelling and simulation of systems here form the promising basis for
a design approach that may ensure optimal systems, when also using the well-established meth-
ods from control engineering, which forms very powerful techniques both in the analysis and
in synthesis of multi-technology (mechatronic) systems. Before control-engineering techniques
may be applied it is, however, required that components and problems may be described (mod-
elled) in terms compatible to the “language” used in control engineering. This means, that the
behaviour of components and systems must be expressed as static and dynamic characteristics,
transfer functions etc. Thus the disciplines of describing the general system dynamics, do-
ing simulation and laboratory measurement are needed for the physical-mathematical dynamic
modelling, efficient handling and solution of the complex describing equations, and experimen-
tal evaluation of models. Finally, all these activities loose most of their sense, if they are not

2
based on a broad knowledge and experience of the intended application, i.e. the environment or
surroundings in which the mechatronic system should work or be a part of.

1.1 Modelling and Simulation as Prerequisites for the Design Process

The control engineering approach has here been legitimated through the above paragraph, and
maybe the most characteristics feature is the use of dynamic models and simulation, which can
be worked out to more or less sophistication depending on the level of analysis relevant to the
actual problem. Hence, it is important to concentrate engineering research and development
efforts in this fields, to establish some of the necessary tools.
Generally, savings in work or cost are the essential reasons why experiments are carried out
with models instead of real systems. However, some situations occur in which no alternative
to model-experiments exist. This is particularly the case when dealing with systems in which
possible instability may initiate irreversibly accelerating processes with hazardous and incalcu-
lable consequences. Other good reasons for using simulation in both analysis and synthesis are
well known and can be stated as:
• The physical system is not available. Often, simulations are used to determine whether a
projected system should ever be built.
• The cost of experimentation is too high. Often, simulations are used where real experi-
ments are too expensive. The necessary measurement tools may not be available or are to
expensive. The system might be used all the time and taking it “off-line” would involve
unacceptable cost.
• The time constants of the system are not compatible with those of the experimenter. Often
simulations are performed because the real experiment executes so quickly that it can
hardly be observed.
• Control variables (disturbances), state variables, and/or system parameters may be inac-
cessible. Often, simulations are performed because they allow us to access all inputs ad
all state variables, whereas in the real system, some inputs may not be accessible to ma-
nipulation and some state variables may nor be accessible to measurement. Simulation
allows us to manipulate the model outside the feasible range of the physical system.
• Suppression of disturbances. Often, simulations are performed because they allow us to
suppress disturbances that are unavoidable in the real system. This allows us to isolate
particular effects, and may lead to better insight into the system behaviour.
• The number of parameters to adjust in the experiment are too large, giving a huge number
of possibilities and at best, result in a sub-optimal solution.
The most important strengths of simulation, but also its most serious drawbacks, are the
generality and ease of its applicability. It does not require much of a genius to be able to utilize
a simulation program. However, in order to use simulation intelligently, (having the handle to
make the real world behave the way we want it to), we must understand what we are doing.
Danger lies in forgetting, that the simulation model is not the real world, but that it represent
the world under a very limited set of experimental conditions.
Simulations are rarely enlightening. In fact, running simulations is very similar to perform-
ing experiments in the lab. We usually need many experiments, before we can draw legitimate

3
conclusions. Correspondingly, we need many simulations before we understand how our model
behaves. While analytical techniques (where they are applicable) often provide an understand-
ing of how a model behaves under arbitrary experimental conditions, one simulation run only
tells us how the model behaves under the one set of experimental conditions applied during
the simulation run. Therefore, while analytical techniques are generally more restricted (they
have a much smaller domain of applicability), they are more powerful where they apply. So,
whenever we have a valid alternative to simulation, we should, by all means, make use of it, i.e.
whenever we have a system with an approximate linear behaviour.
However, as the level of analysis is elevated, as it for instance may happen in connection to
development of new components or new systems, a barrier will appear at a certain level. The
complexity which underlines the development of mechatronic systems should be recognised.
Very few people are capable of making models of real mechatronic systems, and even though
simulation programs are appearing on the market that claims they can simulate pretty much
anything, there are still the number of parameters to adjust and the programs can not design the
system for you.
In the authors view, confidence in the modelling is central to the potential success of any
type of simulation based automated design procedures. This requires that any system should
be build up by sub-structures/models that have well known characteristics, are well understood
and preferably are experimentally verified in some way at an earlier stage. In the following
this approach will be referred to as: PDP = Predictable Design Performance. Hence, system
design should be based on well known sub-systems if the desired reduction in dependency on
prototypes is to be obtained. The main drawback of this is, that according to a performance
index only, PDP will almost certainly never produce the optimal design as compared to what
could be theoretically be obtained using an ideal approach involving development of prototypes,
experimental work and application of computational tools, i.e. FEM and CFD. This approach is
referred to as: ODP = Optimal Design Performance. However, by including development time
or “time to marked” the PDP will be capable of producing the best performance in a certain
time, thereby acting as a kind of “best practice”.

1.2 Automated Design Procedures - Optimisation

Turning our attention to automated design procedures, they are classically referred to as opti-
misation. In textbooks for engineering students, design procedures for e.g. mechanical mech-
anisms, are classically divided into structural and dimensional synthesis although experience
seems to indicate that this division in most cases is somewhat artificial. The reason for this
is that an evaluation of a certain structure/topology cannot be carried out without the dimen-
sional synthesis/sizing unless the structure is clearly unsuitable in the given case. Structural
and dimensional synthesis should be approached as an integrated design task that involves the
determination of both topology and dimensions. Basically, this may be done in two distinctly1
different ways as seen from an algorithmic point of view:
• Serial
• Concurrent
In serial synthesis the approach may roughly be described as follows: Topologies are gen-
erated before hand based on experience or, alternatively, some kind of automated topology gen-
1
Combination of the two methods in multi-level optimisation is possible, see e.g. [21].

4
eration. Next, the actual sizing is carried out, typically by means of some sort of minimisation
technique. Finally, the optimised/sized topologies are compared and the best one is chosen.
In concurrent synthesis the topology and the dimensions are varied simultaneously introduc-
ing a number of challenges with respect to modelling and sensitivity analysis. These challenges
need to be met in such a way that objects, e.g. an actuator or a sub-linkage, may either appear
or disappear as well as having their design parameters varied continuously.
On top of that, both serial and concurrent approaches needs to handle the numerical prob-
lems associated with the mixing of discrete and continuous parameters, since a typical system
is composed of parameters that are to be chosen from standardised values (profile dimensions,
motor sizes, gears with different gear ratios etc.) as well as parameters that may vary freely
(profile length, controller parameters, actuator attachment points etc.).
The use of automated topology generation is further elevated and the design task further
complicated - due to the fact that despite the vast amount of standardised components, the
designer of mechatronic systems can not allow oneself to be overly limited by, e.g. standard
component configurations. Today (special) components or parts are increasingly customized to
meet specific customer demands with respect to functionality and performance.

2 OPTIMISATION BASED DESIGN METHODS

In this section a new paradigm for the design of mechatronic systems is investigated and put
forward - Mechatronic system design being defined as the synergistic integration of mechan-
ical engineering, electronics, intelligent control, and systems thinking in the design of smart
products and processes. It is therefore a very interdisciplinary undertaking, where knowledge
from different domains has to be integrated in an optimal way. It stands apart from Mecha-
tronics understood as an interdisciplinary subject attracting contributions from all related fields
without really putting forward the opportunities and challenges arising specifically due to the
interdisciplinary interactions, see figure 1.

Figure 1: Synergism and integration in mechatronic system design.

The novelty in the idea presented in this paper is to extend the use of algorithm based optimi-
sation to serve as a design tool in mechatronic system design. An algorithm based optimisation
here means that the problem is formulated in terms of an objective function, i.e. a mathemat-

5
ical function describing the design goal(s), that should be minimised and where the allowed
variations in the design variables are described by a number of constraints, like:

minimise f (x) for x ∈ Rn (1)

subject to the constraints:

h (x) = 0
(2)
g (x) ≤ 0
The primary goal is to develop methods, i.e. formulate the optimisation problems, that
are capable of handling both the control architecture and the mechanical system of complex
mechatronic systems. It is in this regard considered essential to the design objective that the
developed methods can operate simultaneously at a concept level and a sizing level while being
capable of including commercially available components.
The results of the project can be used by manufacturers of multi-technological systems and
point out how the integration of disciplines lead to new degrees of freedom in design; it leads
to new research fields and at the same time helps to push research in related fields into new
directions.

2.1 Background and Research areas

A general discussion on inter-disciplinarity in research, its lack in academia and its tremendous
importance for the next century is presented in [1]. Large studies are referred to in which it is
concluded that the technology cannot be partitioned according to conventional disciplines.
Today the engineering of mechatronic systems and products typically implies a subsystem-
based approach [2]. By subsystem-based approach is here meant a product development strat-
egy by which integrated systems are built from technology homogeneous subsystems (actuators,
mechanics, electronics, control and software). The subsystems are developed in a concurrent
manner with an important focus on subsystem interfaces. Once the interfaces are designed,
each subsystem is designed in a fairly traditional way. A typical design scenario is that an early
conceptual design phase results in a coarse partitioning of functionality between mechanics and
control software. Based on this, an initial mechanical design is performed. This design provides
a basis for modelling of system dynamics that in turn provides input to the control design. This
is further developed to real-time software. This means that the focus has been on team building
to improve communication and multidisciplinary understanding between engineers of different
expertise such that the interfaces can be properly defined. In the subsystem-based approach the
performance of the system is a result of a sound integration of existing technology [2]. In the
existing engineering literature on mechatronics, the subsystem-based approach is mostly pre-
dominant, with limited coverage of the development process etc. [3, 4]. Thus, the literature
today does not give an approach for mechatronics engineering.
An interpretation of the above is that state-of-the-art in mechatronics engineering is about
mastering a multitude of disciplines, whereas mechatronics as an engineering science focuses
on interactions among components in different technologies, figure 2.
The strong interaction of technologies (mechanics, control, software, electronics) leads to
strong interdependencies that influence, for example, performance and reliability in a way that
can only be handled and exploited in a real multidisciplinary environment of a substantial size.
Research activities in systems integration should therefore be directed toward operational
systems in order to be really useful from an industrial viewpoint.

6
 Figure 2: Interactions within an mechatronic system.

To overcome these problems, we need a mathematically well-founded theory that integrates

models for the heterogeneous components. Without such a theory, the design will invariably
remain at the sub-system level. The research challenge being:

1. In the process of transferring the functionality from mechanics to control, the idea is typ-
ically to simplify the mechanical design and/or improve the control performance. This
needs new control methods based on better models of the mechanical system. The ideal
mechatronic solution to this problem would be that the deficiency of the mechanical sys-
tem could be cost-effectively compensated for by a suitable control engineering solution.
An example is the development of control methods for systems involving multiple non-
linearities (friction, backlash, saturation, etc.) combined with mechanical flexibility.

2. Another research challenge concerns how to best combine different modelling concepts
in order to support the development of complex industrial systems characterized by: 1)
a combination of event- and time-triggered parts with respect to both requirements and
implementation, 2) a combination of discrete-event and discrete-time and continuous-
time dynamic systems, and requirements on functionality, timing, reliability, safety, etc.

2.2 Research method- Optimisation

Optimisation methods have been used extensively in a number of fields. Within structural and
mechanical system several formulations has been used with great success. In control theory
an entire area is devoted to optimal control considered a mature discipline. The handling of
discrete design variables is also well developed today utilizing different genetic algorithms.
Especially, within the area of fluid power some work has been devoted to the simultaneous
design at concept and sizing level, see for example [5, 6, 7 and 8]. In [5] the use of commercially
available components are introduced by means of databases.
However, to the knowledge of the principal investigators no methods have been developed
with the main purpose of handling the design of mechatronic systems that include all of the
above areas simultaneously. The main obstacle is the combination of the inherent difficulties
mentioned in the second research challenge and, equally important, the needed ability to handle
and evaluate different design concepts. These obstacles are the main reason why optimisation
as a tool is not more widespread in industry today.
Since the mid-nineties the authors have devoted a substantial part of their research to design
and design methods. Initially, emphasis was on the sizing of mechanical mechanisms based on

7
purely mechanical criteria; see [9, 10, 11] although the concept level also has been addressed
[12]. The inclusion of actuation was introduced by sizing of hydraulically actuated manipulator
systems [13, 14, 15, 16], and later, optimal control aspects was included in [17, 18, 19, 20] but
still at a purely sizing level. The concept level on actuation systems was introduced in [21] and
further developed to handle hydraulic-mechanical systems in [22, 23]. Recently, the discrete
design variables and the use of commercially available components have been addressed in
[24] on electrically actuated servo driven robots, and further developed to include a certain
degree of concept level [25] on hydraulically actuated scissor tables. The applied methods have
been optimisation utilizing either gradient or non-gradient based search algorithms or genetic
algorithms whenever discrete design variables have been present. However, in the recent works
[24, 25] some very encouraging new results have been obtained using mapping techniques that
allow the inclusion of discrete design variables, commercially available components and design
variations at concept level simultaneously and, at the same time, avoiding the time consuming
genetic algorithms. These results give confidence that the research challenges of mechatronic
system design can be met successfully.

3 CONCLUSION
This paper has placed its emphasis on integrated modelling and design optimisation. R&D
activities have become more customer-driven than ever before, and therefore simulation tools
for research and development, with strong emphasis on drastically accelerating the product
introduction process has become more and more important. This, in turn, calls for a high degree
of predictability of the initially developed design, with a reduced dependency on prototypes and
an increased dependency on simulation and automated design procedures.

• Only well understood/well described sub systems with a high degree of predictability in
behaviour should be used as building blocks.

• An automated procedure that determines not only the dimensions but also the topology
should be employed using either a serial or a concurrent technique.

In the authors view PDP is the design approach to be used in system design of the future.
Parallel to this, experimental and scientific work will continuously increase the design space,
i.e., the amount of sub systems applicable to the PDP design approach.

REFERENCES
[1] F. Meyer-Kramer. 1997. Science-bases technologies and interdisciplinarity: challenges for
firms and policy. Systems of Innovation, Technologies, Institution and Organizations, C.
Edqvist (ed.), London.
[2] H. van Brussel. 1989. The mechatronics approach to motion control. Proc. Int. Conf. on
Motion Control - The Mechatronics Approach, Antwerp, Belgium.
[3] D. Shetty and R. A. Kolk. 1997. Mechatronics System Design. PWS Publishing Company.
[4] J. R. Hewitt. 1993. Mechatronics - The contributions of advanced control. In Proc. 2nd
Conference on Mechatronics and Robotics, Duisburg, Germany.
[5] J. Andersson and P. Krus. 2001. Multiobjective Optimization of Mixed Variable Design
Problems. Proceedings of the First International Conference, EMO 2001, Zurich, Switzer-
land.

8
[6] A. Jansson, P. Krus and J.-O. Palmberg. 1992. Optimisation of Fluid Power Systems with Two
Alternative Non-Derivative Methods. ASME Annual Winter Meeting, Anaheim, USA.
[7] C. R. Burrows. 1993. Computer tools for the design of fluid power systems. Proc. 3rd Scandinavian
Conference on Fluid Power, Palmberg J.-O. (ed.), Linkping, Sweden.
[8] R. D. Burton and C. M. Sargent. 1989. The use of expert systems in the design of single and
multi-load circuits. International Conference on Fluid Power Transmission and Control, Hangzhou,
China.
[9] M. R. Hansen. 1993. An Automated Procedure for Dimensional Synthesis of Mechanisms. Structural
Optimization, Vol. 5, No. 3, pp. 145-152.
[10] M. R. Hansen and J. M. Hansen. 1999. An Efficient Method for Synthesis of Planar Multibody
Systems Including Shape of Bodies as Design Variables. Journal of Multibody System Dynamics,
Vol. 2, No. 2, pp. 115-143.
[11] M. R. Hansen and T. O. Andersen. 2003. Multi Criteria Design Optimization of Backhoe Loader
Front Mechanism, Int. Mechanical Engineering Congress and R&D Expo, Washington D. C., USA.
[12] M. R. Hansen. 1996. A Multi Level Approach to Synthesis of Planar Mechanisms. Journal of
Nonlinear Dynamics, Vol. 9, pp. 131-146.
[13] M. R. Hansen and T. O. Andersen. 2001. A Design Procedure for Actuator Control Systems Using
Optimization Methods. The 7th Scandinavian International Conference on Fluid Power, SICFP01,
Linkping, Sweden.
[14] M. R. Hansen, T. O. Andersen and F. Conrad. 2001. It-Tool for Design Optimisation of Hydrauli-
cally Actuated Manipulators, The 2nd international Workshop on Computer Software for Design,
Analysis and Control of Fluid Power Systems, Ostrava, Czech Republic.
[15] M. R. Hansen, T. O. Andersen and F. Conrad. 2001. Experimentally Based Analysis and Synthesis
of Hydraulically Actuated Loader Crane, Power Transmission and Motion Control, Bath, UK.
[16] T. O. Andersen and M. R. Hansen. 2004. A Mechatronic Solution for Efficiency Optimization of
Forklift Trucks, REM 5th Int. Conference on Research and Education in Mechatronics, Kielce,
Poland.
[17] M. R. Hansen, T. O. Andersen and F. Conrad. 2002. Control of Oscillations in Hydraulically Driven
Off-Highway Vehicles, Power Transmission and Motion Control (PTMC 2002), Bath, UK.
[18] T. O. Andersen, M. R. Hansen, M. R. and F. Conrad. 2003. Active Damping in Off Road Vehicles,
The Eigth Scandinavian International Conference on Fluid Power, Tampere, Finland.
[19] T. O. Andersen and M. R. Hansen. 2004. Multi Criteria Design Improvement of Commercial Loader
Crane, International Mechanical Engineering Congress and R&D Expo, Anaheim, USA, November.
[20] M. R. Hansen and T. O. Andersen. 2005. Tool point control of mobile hydraulic manipulators.
In Proc. of The Sixth Int. Conference on Fluid Power Transmission and Control (ICFP’2005),
Hangzhou, China.
[21] M. R. Hansen, T. O. Andersen and H. C. Pedersen. 2003. An Approach to Automated System Design
of Hydraulic Systems. Power Transmission and Motion Control (PTMC 2003), Bath, UK.
[22] T. O. Andersen, M. R. Hansen and F. Conrad. 2003. System Topology Optimization on a Mobile
Loader Crane. 18th Int. Conference on Hydraulics and Pneumatics (IHCP2003). Prag, Czech
Republic.
[23] M. R. Hansen, T. O. Andersen and H. C. Pedersen. 2003. System Topology Optimization, 1st Inter-
national Conference on Computational Methods in Fluid Power Technology, Methods for Solving
Practical Problems in Design and Control, Melbourne, Australien.
[24] M. R. Hansen and T. O. Andersen. 2004. Automated Sizing Procedure of Servo-Driven Robot for
Pallettes Handling. International Mechanical Engineering Congress and R&D Expo. Anaheim,
USA.
[25] M. R. Hansen and T. O. Andersen. 2005. Design optimization of hydraulically actuated lifting table
using mapping techniques. In Proceedings of The Ninth Scandinavian International Conference on
Fluid Power, SICFP’05, Linkping, Sweden.