Chapter 2

Actuator Design Analysis

2.1 Nature and Objectives of Actuator Design Analysis

The primary quantities of an actuator depend on

1. the actuator principle or class (e.g. piezoceramic stack actuator, solenoid actua-
tor, hydraulic cylinder);
2. a set of non-geometrical design variables (e.g. the kind of active material used,
the amount of pre-strain of a Shape-Memory wire);
3. a set of geometrical variables;
4. the actuator input quantity;
5. the external load.
While designing an actuator for a particular application, the engineer deals ex-
plicitly or implicitly with this multi-variable dependency. Even with the help of
formal optimization algorithms, the design task can reveal extremely challenging if
no preliminary information is made available to the engineers as a guide to effec-
tively move in the highly multidimensional design space the actuator. With limited
resources for the design procedure, the result is inevitably a sub-optimal actuator
device. Things become even more complex if the actuator is to be designed together
with the host mechanical system.
The scope of this study is to provide such a guide by analyzing the above listed
dependencies in a systematic and general way, discerning strong trends from the
weak ones and nally extracting rules and criteria to be used as a basis for a powerful
and efcient design process.
In particular, those rules and criteria should guide the engineer in:
choosing the proper actuator class or principle for a given application.
optimizing the actuator size and geometry for given output quantities.
perform a simultaneous design of the actuators geometry and of its interface
with a given host mechanical system (in this case the required actuator output
quantities are not dened explicitly but are function of the actuator position and

29
30 2 Actuator Design Analysis

of the direction of the actuator forces, which are among the unknowns of the
problem).
perform a coupled optimization with the external system in which the whole
system is to be concurrently designed.
For some design tasks it can be useful not to refer to the primary output quantities,
but to derived quantities, mostly of energetic nature, like work or power. The work
performed by the actuator against an external load will depend on all the variables
listed above, with the difference that on point d. not a single value of the input
quantity is to be considered but a dened change of the input quantity between a
minimum and a maximum value. The actuator power will be dened by a cyclic
variation of the input quantity as a function of time instead.
The key facts which will be considered in the actuator design analysis presented
in this chapter are discussed in the following:
Effect of the external load. For a given choice of the variables on items 1 to
4, the actuator denes a relationship between the primary output quantities. The
actual value of the actuator force and stroke are only dened if the characteristic
curve of the external load is considered, which provides a second relationship be-
tween those quantities. While considering a parametric choice of possible loads
(e.g. all linear elastic springs, dened by their spring constant) the primary output
quantities can be usually maximized with respect to the load. The same applies
to derived quantities (with the adjustments discussed above). The maximum val-
ues of the considered output quantities are taken into a second step of the design
analysis; the optimal load for a given quantity to be maximized gives a valuable
insight on how actuator and host system have to be adapted to another in order to
exploit the chosen actuation principle in an efcient way.
Thresholds on the input quantities. While considering now the maximum val-
ues resulting by the above described load analysis, these will depend only on the
variables listed on items 1 to 4. The dependency on the input variable chosen to
drive the actuator will usually be of monotonic nature but just up to a thresh-
old value beyond which the input variable cannot be increased. By analyzing the
threshold effects and identifying the limit values the design space can be reduced
by a further dimension.
Geometrical parameterization. The maximum output quantities resulting by
the threshold analysis can now be analyzed as a function of the actuator geom-
etry (for a xed choice of the variables on items 1 and 2). The rst step of the
geometry analysis is the identication of a global representative length which de-
nes the actuator size and of an appropriate number of aspect ratios which relate
the actuator size to a corresponding number of reference lengths along different
axes. The further geometrical variables are then related to the reference lengths
on the same axis by additional non-dimensional variables. In most cases, these
additional variables express the relationship between an active length, area or
quantity of active material to the length, area or quantity of a passive component.
We will therefore use the name lling factors for geometrical ratios dened on
the same axis, implicitly including geometrical ratios of different nature.
2.1 Nature and Objectives of Actuator Design Analysis 31

Size analysis. Since the output quantities are expected to increase indenitely as
a function of the actuator size, the next step in the actuator design analysis is to
identify the law which rules the dependency of the maximum output quantities
on the size and try to separate, if possible, this dependency from the dependency
on aspect ratios and lling factors. If the dependency on size is dened, the anal-
ysis can now be reduced to the study of the dependency of the maximum output
quantities for a given size as a function of aspect ratios, lling factors and nally
the non-geometrical variables listed on items 1 and 2.
Scalability. One important implication of size analysis is the scalability issue. It
deals with the question if the output quantities are mechanically scalable, i.e. if
they change with size like in a passive mechanical system which can be analyzed
by continuum mechanics. This implies forces to scale with the square of size and
strokes linearly with the size. As a consequence, work scales with the cube of
the size. Mechanical scalability is quite relevant for mechatronic and adaptronic
systems, since it allows to globally scale the whole active system consisting of a
passive and an active part.
Shape analysis. Once the output quantities have gone through the successive
steps of load optimization, input quantity threshold and size analysis, the optimal
aspect ratios and lling factors can be found, in order to nally achieve a set of
performance quantities which are representative for the chosen actuator principle
(item 1) and only depend on non-geometrical design variables (item 2).
Actuator principle analysis. This is the nal step of the actuator design anal-
ysis, in which the chosen actuator principle can be analyzed on a general ba-
sis, independently on geometry, thresholds and loads. After having performed
if possible an optimization with respect to the non-geometrical design vari-
ables of item 2, indexes of merit for the chosen actuator principle can be dened
on an objective and quantitative basis, which helps comparing different possible
choices for a given application.
The designer can take advantage from the described design analysis as a whole or
in part, making use of intermediate results and integrating them with conventional
design procedures based on statistical methods or on trial-and-error techniques. In
order to get familiar with the proposed design philosophy, we will summarize it in
form of a stepwise design algorithm which essentially reproduces the successive
reduction of the design space described above.
This algorithm will be applied once the designer chooses a certain class of actu-
ator and denes a rst sketch of the actuator geometry as illustrated in Figure 2.1.
After applying the algorithm, the actuator designer can decide whether the obtained
actuator performance is optimum or not and to introduce or change the actuator class
or geometry and start the procedure again. The procedure can be repeated until it
becomes clear that the best actuator has been chosen for the given application.
The forthcoming sections illustrate the different methodology steps, stressing the
most important concepts which are to become relevant in the design of the actua-
tor. The different step procedures are explained and illustrated by means of exam-
ples. Although the whole methodology is described, it has to be noted that for some
classes of actuators not all the steps will have the same importance. Dimensional
32 2 Actuator Design Analysis

Fig. 2.1 Actuator design procedure

2.2 Performance Indexes 33

analysis, for example, may become valuable in some applications related to uid
power or thermal transfer and may provide trivial results when dealing with other
kind of actuator applications. In any case, it is strongly recommended to follow all
the steps, since some relevant non-expected results may be obtained in certain cases.
Although the present chapter describes different examples to illustrate the method-
ology, a detailed analysis of three classes of conventional actuators (solenoid, mov-
ing coil and hydraulic actuators) will be developed in the following Chapters 3, 4
and 5. The application to solid-state actuators will be the topic of Chapter 6.

2.2 Performance Indexes

Before explaining the actuator design analysis philosophy in more detail, we will
shortly review some published results on actuator performance indexes, which
somehow represent the state of the art on the issue of quantitative comparison of
different actuator class and principles.
As a rule, published performance indexes of actuators are based on statistical
methods, with the exception of solid-state actuators for which a model-based analy-
sis is quite straightforward and customary. With the design analysis described later
in the chapter we aim to provide a novel contribution to this topic by extending
model-based techniques to conventional actuator principles.
The data presented in this section are taken from [5, 4, 11], where the reader can
nd more details on the denition of the used performance indexes and on how the
quantitative results were obtained.
The basic characteristics of actuators are well dened as the stress and the
strain . As stress and strain apply for a broad range of actuator sizes, for a specic
application it is common to use the force F and displacement or stroke x. Stress
and force are linked by the section A as F = A, while strain and displacement are
linked by the reference length l as x = l.
The present Section analyzes and compares different actuator classes according
to different performance criteria. When an actuator has to be chosen for a given
application, the designer has to decide what the relevant criteria are and choose one
actuator or the other according to it.
The analyzed classes of actuators are:
1. Low strain piezoelectric actuators
2. High strain piezoelectric actuators
3. Polymeric piezoelectric actuators
4. Thermal expansion actuators 10 K
5. Thermal expansion actuators 100 K
6. Magnetostrictive actuators
7. Shape memory alloy based actuators
8. Moving coil actuators
9. Solenoid actuators
10. Muscles
34 2 Actuator Design Analysis

11. Pneumatic actuators

12. Hydraulic actuators
The different actuator classes have been compared using different performance
indices. Such indices include:
1. Maximum strain
2. Maximum stress
3. Maximum frequency
4. Maximum volumetric power density
5. Maximum mass power density
6. Efciency
7. Resolution
A plot showing the different maximum strain versus maximum stress is shown
in Figure 2.2. It can be noted that the diagonal lines from left-top to right-bottom
show constant volumetric energy. It implies that a certain actuator can move along
this diagonal by using mechanical amplication or reduction. For example, a high
strain piezoelectric actuator performs strains in the range of 104 . If a mechanical
amplication is added it can increase notably the strain at the cost of reducing the
stress in the same proportion.

3
10

THERMAL EXP. 100 K SHAPE MEMORY

2
MAGNETOSTRICTIVE
10
Maximum Stress max [MPa]

HYDRAULIC
THERMAL EXP. 10 K

1
10
HIGH STRAIN PZ

PZ POLYMER
LOW STRAIN PZ
0
10
PNEUMATIC

MUSCLE
1
10
SOLENOID

MOVING COIL
2
10

3
10 6 5 4 3 2 1 0
10 10 10 10 10 10 10
Maximum Strain max []

Fig. 2.2 Maximum stress versus maximum strain for different classes of actuators (Data extracted
from [5])

Another important quantity is the maximum available frequency. Some actuators

can provide large strains, while others feature high frequencies. Again, mechanical
transmissions can be used in order to increase frequency or strain, but they increase
the volume, weight and cost of actuation system chosen. Figure 2.3 illustrates the
2.2 Performance Indexes 35

8
10

7 LOW STRAIN PZ MAGNETOSTRICTIVE

10 HIGH STRAIN PZ
PZ POLYMER
6
10
Maximum frequency fmax [Hz]

5
10
MOVING COIL
4
10

3
10
HYDRAULIC
MUSCLE
2 PNEUMATIC
10
SOLENOID
1
10
THERMAL EXP. 10 K THERMAL EXP. 100 K SHAPE MEMORY
0
10

1
10

2
10 6 5 4 3 2 1 0
10 10 10 10 10 10 10
Maximum Strain max []

Fig. 2.3 Maximum frequency versus maximum strain for different classes of actuators (Data ex-
tracted from [5])

maximum frequency versus the maximum strain for different classes of actuators. It
can be seen that the actuators featuring the best frequency to strain ratio are mov-
ing coil actuators, muscles, pneumatic actuators, hydraulic actuators and magne-
tostrictive actuators. The thermal expansion actuator 10 K is the worst positioned
according to this index.

8
10
MAGNETOSTRICTIVE
7 LOW STRAIN PZ
10 HIGH STRAIN PZ
PZ POLYMER
6
10
Maximum frequency fmax [Hz]

5
10
MOVING COIL
4
10

3
10
MUSCLE
2 PNEUMATIC HYDRAULIC
10
SOLENOID
1
10
THERMAL EXP. 10 K
SHAPE MEMORY
0 THERMAL EXP. 100 K
10

1
10

2
10 3 2 1 0 1 2 3
10 10 10 10 10 10 10
Maximum Stress max [MPa]

Fig. 2.4 Maximum frequency versus maximum stress for different classes of actuators (Data ex-
tracted from [5])
36 2 Actuator Design Analysis

It can be also useful to see the maximum frequency against the maximum stress,
as shown in Figure 2.4. It can be noted that the actuator featuring a best frequency to
stress ratio is the magnetostrictive actuator, while the solenoid actuator is the worst
positioned according to this index.
The maximum power density is an extremely important quantity since it ex-
presses the maximum power that can be provided per unit of volume or mass. Fig-
ure 2.5 shows the maximum volumetric power density versus maximum strain while
Figure 2.6 shows the maximum mass power density versus maximum strain. Hy-
draulic actuators dominate both volumetric and mass power density to strain ratios,
while thermal expansion 10 K actuators feature the worst power density to strain
ratio.
Multiplying the maximum strain times the maximum stress another interesting
quantity is obtained, which corresponds to the volumetric work. The plot linking
power density and volumetric work of Figure 2.7 shows that the actuator featuring a
best ratio is again the hydraulic actuator, while moving coils, solenoid and thermal
expansion actuators 10 K show the worst ratios.
The resolution versus the maximum strain is illustrated in Figure 2.8. It should
be noted that the resolution is expressed as the minimum controllable strain and
therefore the smaller the value the better the resolution is. It can be seen that the
worst resolutions are obtained by the actuators that provide large strains, like the
muscles or the solenoid actuators, while the best resolution is obtained by low strain
piezoelectric actuators. The diagonal lines bottom-left to up-right show the number
available and controllable positions where the actuator can be moved. The number
of positions is obtained by dividing the strain into the resolution. It can be noted
that the maximum number of positions is obtained by the different piezoelectric
actuators, moving coils and pneumatic and hydraulic actuators, while the thermal
expansion 10 K actuator feature the worst number of positions. Therefore, such
thermal expansion 10 K actuators will be recommended for applications requiring a
low number of positions, such as bistable switches.
The efciency versus the maximum mass power density is illustrated in Fig-
ure 2.9. It can be noted that the actuators with higher efciency are the piezoelectric,
magnetostrictive and hydraulic actuators.

2.3 Design Parameters

The rst step introduces the relevant design parameters in the construction of the ac-
tuator. A detailed schematic drawing is presented showing the geometric quantities
and the materials used. In order to obtain a clear design parameterization geometri-
cal factors, aspect ratios and lling factors are used.
2.3 Design Parameters 37

9
10

Maximum volumetric power density Pmax [W/m ]

3 LOW STRAIN PZ
MAGNETOSTRICTIVE HYDRAULIC
HIGH STRAIN PZ
PZ POLYMER
8
10

SHAPE MEMORY

7
10
THERMAL EXP. 100 K
PNEUMATIC

6 MOVING COIL
10

MUSCLE

5
10
THERMAL EXP. 10 K

SOLENOID

4
10 6 5 4 3 2 1 0
10 10 10 10 10 10 10
Maximum Strain max []

Fig. 2.5 Maximum volumetric power density versus maximum strain for different classes of actu-
ators (Data extracted from [5])

13
10
Maximum mass power density Pmax [W/kg]

MAGNETOSTRICTIVE
LOW STRAIN PZ HIGH STRAIN PZ
12
10
HYDRAULIC
PZ POLYMER SHAPE MEMORY

11
10

THERMAL EXP. 100 K

10
10 MOVING COIL

9
PNEUMATIC
10
MUSCLE
THERMAL EXP. 10 K

8
10 SOLENOID

7
10 6 5 4 3 2 1 0
10 10 10 10 10 10 10
Maximum Strain []
max

Fig. 2.6 Maximum mass power density versus maximum strain for different classes of actuators
(Data extracted from [5])

2.3.1 Geometrical Factors

Geometrical factors dene the ratio between any geometrical dimension and a ref-
erence geometrical dimension in the same axis.
38 2 Actuator Design Analysis

9
10

Maximum volumetric power density Pmax [W/m ]

3 LOW STRAIN PZ HYDRAULIC
MAGNETOSTRICTIVE
HIGH STRAIN PZ
PZ POLYMER
8
10

SHAPE MEMORY

7
10
THERMAL EXP. 100 K
PNEUMATIC

6 MOVING COIL
10
MUSCLE

5
10
THERMAL EXP. 10 K

SOLENOID

4
10 6 5 4 3 2 1 0 1
10 10 10 10 10 10 10 10
Maximum Strain Stress max max [MPa]

Fig. 2.7 Maximum volumetric power density versus maximum strain times maximum stress for
different classes of actuators (Data extracted from [5])

4
10
THERMAL EXP. 10 K THERMAL EXP. 100 K SHAPE MEMORY PNEUMATIC
HYDRAULIC

5
10
MOVING COIL
Resolution min []

6
10
MAGNETOSTRICTIVE

7
10
HIGH STRAIN PZ
PZ POLYMER

8
10
LOW STRAIN PZ

9
10 6 5 4 3 2 1 0
10 10 10 10 10 10 10
Maximum Strain max []

Fig. 2.8 Resolution versus maximum strain for different classes of actuators (Data extracted from
[5])

A non-dimensional geometrical factor ki is obtained for each length li as a quo-

tient of this length and the reference length l in its axis as:

li
ki = li = ki l (2.1)
l
2.3 Design Parameters 39

0
10
SOLENOID MOVING COIL
0
PNEUMATIC 10 LOW STRAIN PZ
MUSCLE
1
10

HIGH STRAIN PZ
HYDRAULIC
SHAPE MEMORY
Efficiency []

Efficiency []
2
10 PZ POLYMER

MAGNETOSTRICTIVE
THERMAL EXP. 100 K
3
10

THERMAL EXP. 10 K
4
10

5
10 7 8 9 10 11 12 11 12 13
10 10 10 10 10 10 10 10 10
Maximum mass power density Pmax [W/kg] Maximum mass power density Pmax [W/kg]

Fig. 2.9 Efciency versus maximum mass power density for different classes of actuators (Data
extracted from [5])

Using these factors, all the lengths in the same axis can be related to one sin-
gle length, simplifying the analysis of the size dependance of different quantities.
The number n of independent reference lengths depend on the degrees of symmetry
of the actuator. An actuator with cylindrical shape presents two different reference
lengths (n = 2, since a cross-section diameter and a length dene a cylinder), a
spherical actuator would be dened with one reference length (n = 1, only a diam-
eter denes a sphere).

Example 2.1. Evaluate the mass of a hollow sphere.

A hollow sphere can be characterized by the outer diameter D2 and the inner
diameter D1 . Considering the material density the mass msphere of the sphere can
be expressed as:

msphere = (D32 D31 ) (2.2)
6
The previously mentioned number of independent reference lengths is n = 1 in
this case. Using the geometrical factor k1 = D1 /D2 the dependance on D1 can be
removed, allowing to express (2.2) as

msphere = D32 (1 k13 ) (2.3)
6
40 2 Actuator Design Analysis

2.3.2 Aspect Ratios

The proportions between different axes can be characterized by means of aspect

ratios.
Aspect ratios express the relationship between different reference lengths as:
i
= (2.4)

where i refers to the reference length corresponding to the axis i, while corre-
sponds to the absolute reference length, as a function of whom all the other geomet-
ric quantities will be expressed.
For the cylindrical coordinates case, it yields:
r
= (2.5)
l
where r corresponds to the main radial reference length and l corresponds to the
main axial reference length.
If n independent reference dimensions are necessary, n 1 aspect ratios are to be
dened. The combination of the previous two concepts implies that all the geomet-
ric dimensions are expressed as a function of one single reference length, which is
associated to the size of the actuator and allows the independent study of the perfor-
mance of an actuator with a limited size and the actuator performance when the size
is changed.

Example 2.2. Evaluate the weight of the pipe illustrated in Figure 2.10.

Fig. 2.10 Pipe analyzed in

Example 2.2

The pipe can be characterized by the geometrical quantities D1 , D2 and L. Con-

sidering the material density the mass m pipe of the pipe can be expressed as:

(D22 D21 )L
m pipe = (2.6)
4
2.3 Design Parameters 41

The previously mentioned number of independent reference lengths is n = 2 in

this case. Using the geometrical factor k1 = D1 /D2 the dependance on D1 can be
removed. As we need n 1 = 1 aspect ratios to express the weight as a function of
a single geometrical quantity, an aspect ratio = D2 /L can be dened, allowing to
express (2.6) as
(1 k12 )
m pipe = D32 (2.7)
4
where it can be noted that the pipe mass is expressed as a function of a single
length D2 , a material property , the dened geometrical relationship k1 and the
aspect ratio .

2.3.3 Filling Factors

The lling factor provides the portion of usable material against the total needed
material. This is of great importance when dealing with multiple conductors, since
they usually need isolation from the surrounding conductors losing also some space
due to construction issues or the design of the actuator. Such multiple conductors can
be found in a broad range of actuators technologies, from the electrical conductors
employed in an inductance or an electromagnet to the uid conductors of pneumatic
or hydraulic systems. In the mentioned cases the entire cross-section designed for
the conductors wires is not employed.
Filling factors express the relationship between the usable area and the total area
as:
Suse
kf f = (2.8)
Stotal
where Suse is the usable cross-section, Stotal is the total cross-section and k f f de-
pends not only on the isolators width but in the distribution of the wires or conduc-
tors. If squared or rectangular conductors are used, the most optimum lling factors
are obtained, since no space is lost as it happens with circular or elliptic conductors.
For such squared conductors the ratio between the usable surface of D1 D1 and
the total surface of D2 D2 would be k f f = k12 , being k1 = D1 /D2 .
For the example of an electrical coil illustrated in Figures 2.11 and 2.12, Suse =
Scopper is the copper section and Stotal the overall cross-section of the coil. The lling
factor depends on the conguration of the wires in the available space.
If the conguration of Figure 2.11 is used, the lling factor can be dened as

Scopper mnD21 /4 2
kf f = = = k1 (2.9)
Stotal mnD22 4

Using the conguration of Figure 2.12, Stotal can be expressed as

42 2 Actuator Design Analysis

Fig. 2.11 Coil wiring scheme employed in the k f f calculation developed in (2.9)

Fig. 2.12 Coil wiring scheme employed in the k f f calculation developed in (2.11)

    
1    1 n1
Stotal = D2 m + D2 + D2 (n 1) sin = D22 m + 1+ 3
2 3 2 2
(2.10)
for n > 1 and m > 1.
The lling factor

Scopper mnD21 /4 mn
kf f = = 2  =    k12 (2.11)
Stotal D2 m + 2 1 + 2 3
1 n1
4 m + 2 1 + n1
1
2 3

In order to compare the different congurations regardless coefcient k1 , the

previously dened lling factors can be expressed as k f f i = i k12 . Comparing the i
parameters for squared conductors (i = 0), circular placed as in Figure 2.11 (i = 1)
and circular placed as in Figure 2.12 (i = 2), the results displayed in Figure 2.13
are obtained, where it can be noted that squared conductors do not loose any space
due to their conguration and the conguration sketched in Figure 2.12 performs
better than that of Figure 2.11 when the number of conductors is large enough. It
can also be observed that for i = 0 0 = 1 and for i = 1 1 = /4 0.7854. For
i = 2 2 isa monotonically increasing function, with an horizontal asymptote at
3 = /(2 3) 0.9069.
2.4 Output Quantities 43

1.1

0
1
1

2

0.9

0.8

0.7

0.6

0.5

0.4 0 1 2
10 10 10
m,n with m=n

Fig. 2.13 Comparison of the i parameters of the different exposed conductor congurations

Although it has been shown that the lling factor depends signicantly on the
wiring conguration coefcient , it is usually considered as a constant known value
between 0.75 and 0.8.

2.4 Output Quantities

An actuator can be dened as an energy converter which transforms energy from

an external source into mechanical energy in a controllable way. The actuator input
quantities depend on the type of energy used and include all the quantities involved
in the energy conversion from the energy source to the output mechanical quan-
tities. For electromagnetic, piezoelectric and magnetostrictive actuators the input
quantities can be the current, the charge or the voltage; for uid power actuators the
uid pressure or the ow; for shape memory alloys and thermal expansion actuators
the temperature. The output quantities are the relevant mechanical quantities force,
stroke, work or power.

2.4.1 Output Quantities Expression

The output quantities developed by an actuator can be controlled by modifying the

input quantities described in Table 2.1. Such input quantities are provided by a con-
trol system which lead output quantities to the referenced values. Such quantities
are ruled by the mechanical load system or structure, which denes the relationship
44 2 Actuator Design Analysis

between the force and the stroke. The integration of actuators and loads in a mecha-
tronic or adaptronic1 system allows the conception of a unique system which is to
be analyzed.

Table 2.1 Actuator drive input quantities for different actuator technologies
Actuator technology Actuator drive input quantity

Electromagnetic actuators Electrical voltages and currents

Electrostatic actuators Electrical voltages and currents
Hydraulic actuators Oil pressure
Pneumatic actuators Air pressure
Thermal expansion actuators Temperature
Piezoelectric actuators Electrical voltage and charge
Magnetostrictive actuators Electrical voltages and currents
Magnetorheological actuators Electrical voltages and currents
Electrorheological actuators Electrical voltages and currents
Shape memory alloy actuators Temperature

A general expression of output quantities as a function of all input quantities

involved is developed in the second step. These expressions are taken from the gen-
eral physics laws ruling the actuators concerned. Each type of actuator behaves in a
different way and its expressions are presented describing all the assumptions done.
If the load is considered, the different steady-state working points depending on the
load can be discussed and shown graphically.

Example 2.3. Evaluate the force produced by a hollow cube submerged in water
sketched in Figure 2.14.

Fig. 2.14 Sketch of the sys-

tem under analysis in Exam-
ple 2.3

The force produced by the cube can be expressed as the force due to submerging
it into water minus the weight of the cube as
 
Fcube = water D21 xg cube D31 D32 g (2.12)

1 Adaptronics is a term referred to the analysis, design and integration of smart structures and
systems.
2.4 Output Quantities 45

where g 9.81 m/s2 is the gravity constant . Using the geometric factor k1 =
D2 /D1 and the relative cube density = cube /water
 
x  
Fcube = water D1 g
3
1 k1
3
(2.13)
D1
where it can be noted that the maximum force is produced  for x D1 when
the cube is absolutely submerged Fcubemax = water D31 g 1 1 k13 and the
minimum force is produced for x = 0 and yields Fcubemin = water D31 g 1 k13
as shown in Figure 2.15. Between
 these limits the force varies linearly, being exactly
equal to zero for x = D1 1 k13 .

20
Fcubemax

15

10
Fcube

Fcubemin
10

0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

x

Fig. 2.15 Cube force characteristic of Example 2.3

2.4.2 Steady-State Analysis

The output quantities analysis can be addressed by considering static or steady-state

conditions of the system or considering the system dynamics. Depending on the
necessary analysis different approaches have to be taken. The steady-state analysis
provide useful information about steady operating points, where the actuator and
the load nd the equilibrium. The steady-state analysis does not consider how the
system moves from one equilibrium to another, but it provides very important infor-
mation to be employed in the design of the actuator. The present section considers
steady-state analysis. Some considerations on actuators dynamics are exposed in
Section 2.10
46 2 Actuator Design Analysis

In order to establish the characterization plot of the actuator, force-stroke curves

and work-stroke curves can be used, taking into account the output quantities as a
function of the displacement or stroke of the actuator. The inputs (currents, voltages,
pressures, etc.) capable of changing these curves are presented, explaining why and
how they can inuence the actuator performance. The design parameters cannot be
considered inputs and their inuence is discussed in following steps.

Example 2.4. Capacitive actuators (Figure 2.16) are a class of electrostatic MEMS
actuators. Obtain the force-stroke and work-stoke curves for such actuators.

Fig. 2.16 Capacitive actuator

of Example 2.4
z

F
A

The energy stored EC by a capacitor of capacitance C yields EC = (1/2)CV 2

where V is the applied voltage. Considering two opposite parallel plates, the capac-
itance C can be expressed as C = A/z where is the dielectric permittivity , A is
the plate surface and z the distance between plates. The force between plates yields:

AV 2
FC = (2.14)
2z2
The work can be expressed as:

AV 2
WC = (2.15)
2z
Force-stroke and work-stroke curves are plotted in Figures 2.17 and 2.18.

Example 2.5. Considering the capacitive electrostatic actuator of Example 2.4, dis-
cuss the evolution of the equilibrium points when changing the voltage if the actua-
tor is attached to an elastic load.
The actuator and load curves are dened by

AV 2
FCactuator = (2.16)
2z2
2.4 Output Quantities 47

4
10
V=10 V
V=20 V
V=30 V
2
10 V=40 V
V=50 V

0
10
FC [N]

2
10

4
10

6
10
0 1 2 3 4 5 6 7 8 9 10
x [ m]

Fig. 2.17 Force-stroke curve of Example 2.4

5
10
V=10 V
V=20 V
6 V=30 V
10
V=40 V
V=50 V
7
10
WC [J]

8
10

9
10

10
10

11
10
0 1 2 3 4 5 6 7 8 9 10
x [ m]

Fig. 2.18 Work-stroke curve of Example 2.4

FCload = kLoad z (2.17)

The equilibrium if found when FCactuator = FCload , that is

AV 2
zeq = 3 (2.18)
2kLoad

3 AV 2 kLoad
2
FCloadeq = FCactuatoreq = (2.19)
2
48 2 Actuator Design Analysis

Equilibrium points are plotted in Figure 2.19 using the values = 8.854 1012
F/m, A = 106 m2 , kLoad = 3 105 N/m and voltage varying from 10 to 50 V.

0.2
V=10 V
0.18 V=20 V
V=30 V
0.16 V=40 V
V=50 V
0.14
Load
0.12
FC [N]

0.1 V=50 V
V=40 V
0.08
V=30 V
0.06
V=20 V
0.04
V=10 V
0.02

0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
x [ m]

Fig. 2.19 Force-stroke curve equilibrium points of Example 2.5

2.5 Thresholds

Some physical limits (maximum allowed temperature, mechanical resistance etc.)

do not allow the actuator output quantities to be increased indenitely. Since the
purpose is to separately deal with the maximum force, stroke and work available in
a given size and the performance scalability the allowed quantities and parameters
are:
geometric quantities (reference lengths)
geometric relationships (geometrical factors, aspect ratios and lling factors)
material properties (magnetic permeability, resistivity, temperature coefcient,
conductivity, etc.)
universal physics constants (0 , 0 , etc.)
physical thresholds (maximum temperature, stress, etc.)
Therefore, all the other quantities (currents, magnetic uxes, pressures, etc.) must
be expressed as functions of the mentioned quantities. The physical thresholds limit-
ing the maximum force, stroke and work must be analyzed, showing how they limit
the performance of the actuator.
The principle limiting quantities in some common actuators are shown in Ta-
ble 2.2.
2.5 Thresholds 49

Table 2.2 Actuator limiting quantities

Actuator technology Actuator limiting quantities

Electromagnetic actuators Temperature, magnetic saturation, mechanical resistance

Electrostatic actuators Electrical eld, mechanical resistance, temperature
Hydraulic actuators Mechanical resistance, uid losses, temperature
Pneumatic actuators Mechanical resistance, uid losses, temperature
Thermal expansion actuators Mechanical resistance, temperature
Piezoelectric actuators Electrical eld, mechanical resistance, temperature
Magnetostrictive actuators Magnetic saturation, temperature, mechanical resistance
Magnetorheological actuators Magnetic saturation, temperature, mechanical resistance
Electrorheological actuators Electrical eld, temperature, mechanical resistance
Shape memory alloy actuators Mechanical resistance, temperature

Example 2.6. Discuss the limiting electrical limiting quantities involved in the elec-
trostatic capacitive actuators studied in Example 2.4.
The maximum applicable voltage Vmax depends on the maximum allowed elec-
trical eld Emax and the displacement z as

Vmax = Emax z (2.20)

Therefore, the more the distance between plates is increased the higher the volt-
age that can be applied. A reasonable example value for Emax could be around
1000 V /mm, where it can be noted that this class of actuator may be specially ade-
quate in the micro-range where a voltage of 1 V would correspond to a displacement
of 1 m.

Example 2.7. Analyze the ux owing in a toroidal inductance and discuss the max-
imum ux considering the magnetic saturation as the limiting quantity.
The magnetization curve (Figure 2.20) show the relationship between the ux
density B and the magnetic eld intensity H. It is common to use the expression

B = H = r 0 H (2.21)

where is the magnetic permeability obtained as the product of the rela-

tive permeability r characteristic of each material and the vacuum permeability
0 = 4107 NA2 . If there was no saturation, would be constant regardless the
value of H. In practice, this is only the case for operation at low values of H, for
higher values saturation effects become more important and cannot be considered
constant anymore.
Nonlinearities linking B and H have been deeply analyzed in the last decades [17,
15, 14, 13, 12, 6], considering saturation and hysteresis limit cycles. In this example
a very simplied expression is employed, since the purpose of the example is to
illustrate the methodology rather than obtain exact expressions. Expression (2.21)
can be for example expressed as
50 2 Actuator Design Analysis

0.14
n=3
n=3.5
0.12 n=4

0.1

0.08
B [T]

0.06

0.04

0.02

0
0 50 100 150 200 250 300 350 400
H [A/m]

Fig. 2.20 BH saturation curve of Example 2.7

 1
Hn n
B= 0 rH k1 (2.22)
k0 + H n

where rH is a constant that corresponds to the value of r for H = 0.

The magnetic eld intensity H can be dened as H = Ni/l, where N is the num-
ber of turns of the coil and i is the current owing in the coil. In order to pre-
vent saturation, a threshold Hmax has to be dened, implying a limit in the current
imax = Hmax /N. The magnetic ux owing in the coil can be expressed as = B/S
where S is the cross-section. The maximum ux will be given by
 1
0 rH k1 n
Hmax n
max = (2.23)
S k0 + Hmax
n

Similar expressions could be derived using an alternative formulation which

links the ux to the magnetomotive force F by dening the magnetic reluctance
= F/ . This alternative formulation allows to analyze such a system as an elec-
trical circuit , where ux corresponds to electrical current, magnetomotive force
corresponds to voltage and reluctance corresponds to resistance.

2.6 Maximum Target Quantity for a Given Size

The limiting quantities expressions developed in the latter section, can be substituted
in the target quantities expressions, allowing them to be maximized for the given
2.6 Maximum Target Quantity for a Given Size 51

limiting values and design parameters. Such target quantities can range from the
output mechanical quantities to other relevant quantities.

2.6.1 Output Mechanical Quantities Maximization

The analysis of a given output quantity, for a given class of actuators, as a func-
tion of the design variables, shows (as a rule) a monotonic dependance on certain
size variables. For instance, the limit force will typically depend monotonically on
the actuator cross section and the actuator stroke on the length. Other design vari-
ables will exist (typically some aspect or shape factors) for which the considered
output quantity can be maximized. The search of such optimal design variables is
very useful, since it enables a fair and effective comparison between actuators of
different kinds and can be of essential importance for the optimization of the whole
active mechanical system. These expressions are carefully analyzed keeping the size
constant.
The limit force, stroke and work available in a given size are studied depending
on the different design parameters. The geometric factors, aspect ratios and the ma-
terials selected providing the best performance are discussed. This leads to a general
expression of the maximum force, stroke and work in a given volume and provides
design rules to optimize the actuators performance with the proper ratios and ma-
terials.
The output quantity to be maximized can be expressed as a function of a refer-
ence length L, limiting quantities 1 , 2 , ..., universal constants, material properties
1 , 2 , ..., geometrical factors k1 , k2 , ..., kn , aspect ratios 1 , 2 , ...m , lling factors
kf f .
In many cases it is possible to separate the different kind of terms appearing in the
output quantity expression, allowing a separated analysis of the different terms.
Geometrical factors, aspect ratios and lling factors can be separated from other
involved quantities, dening a design factor. The maximization of such a design
factor implies the maximization of the concerned output quantity by means of an
appropriate actuator design.
Moreover, with the purpose of separating all the different factors involved in the
output quantities expression the following functions can be dened:
 kuc
Universal constants 
Design factor d f k1 , k2 , ..., kn , 1 , 2 , ...m , k f f
Geometric function L (L)
Limiting quantities function lim (1 , 2 , ...)
Material properties function mat (1 , 2 , ...)
If it is possible to algebraically separate the different previous functions, the se-
lected output quantity to be optimized can be represented by
 
max = f kuc , d f , L , lim , mat (2.24)
52 2 Actuator Design Analysis

In a number of practical cases the previous functions are easily separable and it is
possible to express (2.24) as

max = kuc L lim mat (2.25)

which has the straightforward property of allowing to maximize or minimize

by maximizing or minimizing all the terms separately.
Example 2.8. Discuss the maximum force and work in the electrostatic capacitive
actuator studied in Examples 2.4 and 2.6.
The previously developed maximum voltage (2.20) can be substituted in (2.16).
The maximum output force yields

AEmax
2
FCmax = (2.26)
2
which is the type of (2.25) with:
kuc = 1/2
d f = 1
L = A
lim = Emax
2 with = E
1 max
mat =
The maximum work:
AzEmax
2
WCmax = (2.27)
2
which is again the type of (2.25) with:
kuc = 1/2
d f = 1
= Az
lim = Emax
2 with = E
1 max
mat =
Example 2.9. A given actuator force is expressed as
 
F = APmax k12 1 k12 ( + 1) (2.28)
where Pmax is the maximum limiting pressure, A is the actuator cross-section, k1
is a geometrical factor and is an aspect ratio. Discuss the maximum force.
The force expression is similar to (2.25), therefore it will be enough to maximize
all the separate terms to maximize the force. Actually, the only non-constant terms
are those included in the design factor d f , which can be dened as:
 
d f = k12 1 k12 ( + 1) (2.29)

Analyzing the previous expression a maximum can be found for k1max = 2/2,
max = 1/2 and d f max = 1/16 as illustrated in Figure 2.21. The maximum force
yields Fmax = APmax /16.
2.6 Maximum Target Quantity for a Given Size 53

Design factor

Maximum kdf=1/16
0.08

0.07

0.06

0.05

0.04

0.03

0.02

0.01

0
1
0.8 1
0.6 0.8
0.4 0.6
k1 0.4
0.2 0.2
0 0

Fig. 2.21 Design factor of Example 2.9

2.6.2 Other Quantities

Other target quantities can be considered instead of the mentioned mechanical out-
put quantities. It may be interesting to address the minimization of the actuator cost
or other indices related to economics, maintenance, control or environmental issues.
Moreover, in many cases it can be specially useful to study some cost to mechanical
quantities ratios, such cost/energy [C/J].

Example 2.10. Discuss the minimum cost/energy ratio for the actuator studied in
Examples 2.4, 2.6 and 2.8.
The actuator cost c can be considered proportional to the n-power of the actuator
volume as cE = kc An zn , where kc is the proportionality constant between size and
cost. The energy expression (2.15) can be employed to compute the cost/energy
ratio as
cE kc An zn 2kc An1 zn+1
= = = (2.30)
WC AV 2 V 2
2z
Using the maximum voltage from (2.20)

2kc An1 zn1

= (2.31)
Emax
2

where it can be noted that the higher the maximum allowed electrical eld Emax or
the electrical permittivity the lower the ratio . Obviously, the ratio can also
be minimized by reducing the parameter kc . As long as the volume dependance is
concerned, this is to be discussed in the following step, which considers the target
quantities as functions of the size.
54 2 Actuator Design Analysis

2.7 Scalability

In the fth step the actuator performance as a function of the size is analyzed and the
scalability and application range are discussed. If two geometrically similar passive
mechanical systems with a given scale factor are considered, and their mechanical
behavior can be described by an approach based on continuum mechanics, they are
to be loaded by forces whose ratio is the square of the scale factor in order to produce
the same stress and strain distribution and, consequently, a similar displacement
eld.
A certain actuator class is mechanically scalable if its output quantities follow
the same rules, i.e. if (by proportional scaling in all directions) the actuator force is
proportional to the square of the size and the actuator displacement is proportional
to the size. Usually, a certain actuator class will be mechanically scalable only in a
certain size range; beyond this range, the required actuator size changes with respect
to the rest of the mechanical system, which can make the use of the considered type
of actuator unpractical.
If the force is proportional to the area and the stroke to the length in the whole
actuators domain it follows that the work is proportional to the volume since the
work is obtained from the integration of the force between two different strokes. If
all these requirements are fullled and the actuator is fully scalable signicant con-
sequences arise. The structures working in the elastic region are considered fully
scalable loads, the expression ruling its behavior = E shows that the quotient
between the stress = F/A and the strain = x/L is the Young Modulus E without
dependance on the size of the structure. Therefore, if the actuator is shown to be
scalable and the load is scalable in the sense described above, the whole system (ac-
tuator plus load structure) would be scalable, allowing the development of models
of easy (normal size) construction as a preliminary step to the construction of large
or small systems (inside the scalability range), with the corresponding saving of re-
sources. In this case, the experimental results must be analyzed as non-dimensional
numbers and provide information for all the range of sizes where the scalability can
be assumed.
Scalability analysis is based on certain assumptions. Since the described assump-
tions are necessary to consider one actuator scalable, the non-scalability when these
assumptions do not hold true has to be also studied.

Example 2.11. The force of a certain actuator is expressed as Fa = kr , where is a

constant > 0. Discuss the scalability of the actuator as a function of the parameter
.
It is straightforward to note that the actuator will be scalable in the sense de-
scribed above as long as = 2. In such a case the actuator force per cross-section
will be constant Fa /r2 = k. Considering the general case, the force per cross sec-
tion yields
Fa
= kr2 (2.32)
r2
2.8 Dimensional Analysis 55

The force per cross section performance as a function of the parameter is plotted
in Figure 2.22, where it is shown that for = 2 the force is clearly scalable. For larger
values, the force per cross section is increased with the size, recommending the
use of such actuators for large actuators. On the other side, when < 2, the actuator
performance improvement is decreased when the size is increased, recommending
its use for applications demanding small actuators.

10
10
=1
=1.5
=2
5
10 =2.5
=3
Fa/A [N/m ]
2

0
10

5
10

10
10 10 5 0 5 10
10 10 10 10 10
r [m]

Fig. 2.22 Force per cross section performance as a function of the parameter

2.8 Dimensional Analysis

It is not always possible to use analytical methods to derive the expressions ruling
actuators. Furthermore, even when analytical methods are feasible it can be conve-
nient to validate such results by means of simulations and experimental validation.
Such validations may be complemented (or in some cases even substituted) with the
so-called dimensional analysis techniques.

2.8.1 The Buckingham Pi Theorem

Dimensional analysis [18] is based on considering dimensionless groups of quan-

tities to form the analyzed expressions. Such analysis allows to obtain some rela-
tionships that can become a valuable complement to other developed analysis and
experimentation.
56 2 Actuator Design Analysis

The dimension of a certain quantity u can be expressed as [u]. All the quantities
can be expressed in some systems containing the basic reference dimensions. The
most common systems are:
Mass-Length-Time MLT
Force-Length-Time FLT
Mass-Length-Time-Temperature MLT
However, the reference system can be chosen depending on each application. Some
usual quantities dimensions can be expressed depending on the reference system as:
Length, in both MLT and FLT systems yields [l] = L
Time,in both MLT and FLT systems yields [t] = T
Mass, in MLT system [m] = M, in FLT system [m] = FL1 T 2
Force, in MLT system [F] = MLT 2 , in FLT system [F] = F
Volume, in both MLT and FLT systems yields [V ] = L3
Acceleration, in both MLT and FLT systems yields [a] = LT 2
The so-called Buckingham Pi Theorem [1] proposes a method of forming dimen-
sionless groups to characterize a certain system. If an equation f (u1 , u2 , ...um ) = 0
is characterized by m quantities ui , i = 1..m, it can be reduced to an equation with
m n dimensionless groups where n is the minimum number of reference dimen-
sions required to characterize the system. The m n dimensionless groups satisfy

(1 , 2 , ...mn ) = 0 (2.33)

The Buckingham Pi Theorem can be applied according to the following steps:

1. List the m quantities involved in the analyzed expression.
2. Dene a reference system of n dimensions (MLT , FLT ,...) and express the
quantities dimensions in such a reference system.
3. Construct a matrix, having n rows and m columns, where the components ai j
correspond to the index of dimension j corresponding to the quantity i.
4. Find the m n dimensionless groups solving the system m i=0 ai j = 0, j = 1..n
5. Using expression (2.33) substitute the m n dimensionless groups.

Example 2.12. Analyze the force produced by a uid power actuator using dimen-
sional analysis considering only the force, pressure and area.
The quantities involved can be listed and expressed in the FL system:
Force [F] = F
Pressure [P] = FL2
Area [A] = L2
The dimension matrix yields:

F P A
F 1 1 0 (2.34)
L 0 2 2
2.8 Dimensional Analysis 57

Considering F = KPa Ab the following equations are obtained

1=a (2.35)
0 = 2a + 2b (2.36)

obtaining a = 1 and a = b = 1. Thus

F = KPA (2.37)

which matches perfectly with the well known force expression if K = 1.

Example 2.13. Analyze the force of an electrostatic capacitive actuator using dimen-
sional analysis.
The quantities involved can be listed and expressed in the FLV system:
Force (N) [F] = F
Voltage (V) [V ] = V
Dielectric permittivity (C2 N1 m2 =N V2 ) [] = FV 2
Area (m2 ) [A] = L2
Distance (m) [z] = L
The dimension matrix yields:

FV Az
F 1 0 1 00
(2.38)
L 0 0 0 21
V 0 1 2 00

Considering F = KV a b Ac zd the following equations are obtained

1= b b= 1
0 = 2c + d d = 2c (2.39)
0 = a 2b a= 2

Substituting in the previous force expression

F = KV 2 Ac z2c (2.40)

Therefore, the force can be expressed as

 
A
F = KV 2 (2.41)
z2

AV 2
which matches clearly with the well known expression FCactuator = 2z2
, using
K = 1/2 and (u) = u.
Example 2.14. Analyze the force performed by a pneumatic actuator.
Before facing the dimensional analysis problem, some important concepts have
to dened:
58 2 Actuator Design Analysis

The orice ow expression of a gas entering in a chamber yields

P
m = Q = KC0 A (2.42)
T
where m is the mass ow, Q is the volumetric ow, the gas density, A is the
orice area, P the gas pressure, T the gas temperature and K and C0 are constants.
The ideal gases constant yields
P
R= (2.43)
T

The adiabatic process constant taking place in the cylinder can be expressed as

K pt = PT 1 (2.44)

where = CP /CV being CP the specic heat coefcient for constant pressure and
CV is the heat coefcient for constant volume [2].
The orice constant can be dened


  +1

 2 1
C0 = (2.45)
R +1

It is important to note that the previous expressions are not known before begin-
ning the dimensional analysis. The quantities involved in the dimensional analysis
can be listed and expressed in the FLT system:
Force (N) [F] = F
Position (m) [x] = L
Pressure (Nm2 ) [P] = FL2
Density (Ns2 m4 ) [] = FT 2 L4
Flow (m3 s1 ) [Q] = L3 T 1
Temperature (K) [T ] =
Ideal gases constant (m2 s2 K1 ) [R] = L2 T 2 1

Adiabatic process constant (K 1 N m2 ) [K pt ] = FL2 T 1
Orice constant (sK1/2 m1 ) [C0 ] = T L1 1/2
The dimension matrix yields:

F x P Q T R K pt C0
F 1 0 1 1 0 0 0 1 0
L 0 1 2 4 3 0 2 2 1 (2.46)
T 0 0 0 2 1 0 2 0 1

0 0 0 0 0 1 1 1 1/2
2.8 Dimensional Analysis 59

g h
Considering F = Kxa Pb c Qd T e R f K pt C0 the following equations are obtained

1= b+c+g b = 1 d/2 f + h/2 g

0 = a 2b 4c + 3d + 2 f 2g h a= 2 2d
(2.47)
0= +2c d 2 f + h c= d/2 + f h/2

0= e f + 1 g + h/2 e= f g 1 h/2

Therefore, the force can be expressed as

f g 1 h/2
F = Kx22d P1d/2 f +h/2g d/2+ f h/2 Qd T g h
R f K pt C0 (2.48)

Rearranging, the following force expression using the different non-dimensional

groups can be found:

F RT K pt C0 P
= K , , , (2.49)
x2 P x2 P P PT 1 T

Q RT
obtaining the non-dimensional numbers 0 = F
x2 P
, 1 = 2
x P
, 2 = P , 3 =

K pt C0 P
and 4 = T
.
PT 1
The following set of non-dimensional numbers can be dened:
a = 0 = xF2 P represents the well known relationship between pressure and
force F = PA.
b = 1 41 = CQ PxT2 characterizes the orice expression (2.42).
0
c = 2 = RT
P corresponds to the ideal gases law expressed in (2.43).
K
d = 3 = pt matches the adiabatic process expression (2.44).
PT 1
1/2
e = 4 2 = C0 R which corresponds to the orice constant (2.45).
The nal expression can be represented as

F Q T RT K pt
= K , , ,C0 R (2.50)
x2 P C0 Px2 P PT 1

2.8.2 Non-Dimensional Numbers

Non-dimensional numbers are useful to characterize certain behaviors that are not
easily established by a given quantity. Some non-dimensional numbers are well es-
tablished in different engineering disciplines, ranging from thermal to uid dynam-
ics engineering. Representative non-dimensional numbers employed in uid dynam-
ics are:
60 2 Actuator Design Analysis

The Reynolds number denes the relationship between the inertial and viscous
forces in a moving uid. It yields

vd
Re = (2.51)

where is the uid density, v is the uid velocity, d is signicant length and is
the absolute uid viscosity.
The Euler number characterized the ratio between pressure and inertia forces in
a uid. It yields
2P
Eu = 2 (2.52)
v
where P is the pressure, is the uid density and v is the uid velocity.
Some well-known thermal transfer non-dimensional numbers:
The Nusselt number denes forced convection, being established as

hc L
Nu = (2.53)

where hc is the convection coefcient, L is a characteristic length and is the
thermal conductivity of the uid producing the forced convection.
The Fourier number quanties heat conduction. It yields
t
Fo = (2.54)
R2
where is the thermal diffusivity, t a characteristic time and R the physical length
where conduction is being produced.
The Grashof number characterizes free convection. It yields

g (Ts T ) L3
Gr = (2.55)
2
where g is the gravity acceleration, the volumetric thermal expansion coef-
cient, Ts the source temperature, T the quiescent temperature, the kinematic
viscosity and L a characteristic length.
The Prandtl number quanties the relationship between viscous and thermal dif-
fusion rate. It is usually expressed as
cp
Pr = = (2.56)

where c p is the specic heat, is the viscosity, the thermal conductivity, is
kinematic viscosity and is the thermal diffusivity.
The Rayleigh number establishes the ratio between buoyancy and viscous forces
in free convection. It yields
2.9 Validation 61

g (Ts T ) L3
Ra = (2.57)

satisfying Ra = GrPr
Other well-known non-dimensional numbers are:
The Poisson ratio is employed in mechanical engineering and links the defor-
mations produced in different axes as
x
= (2.58)
y

The power factor cos is employed in electrical engineering and denes the re-
lationship between active P power and reactive Q power in an alternating current
circuit as
P
cos =  (2.59)
P2 + Q2

2.9 Validation

The seventh and nal methodology step validates the obtained results by means
of prototyping and/or simulation and/or comparison with industrial devices. This
will usually allow the designer to decide whether the actuator is the optimum or
if it is convenient to introduce new modications and to start again all the design
procedure.

2.9.1 Prototype Construction

The classical approach of results validation is based on the construction of a proto-

type and the validation of the theoretically developed expressions with such a real
prototype. The scalability of the actuator discussed in Step 5 can be used to build
scale actuators and structures and generalize conclusions for different sizes.

2.9.2 Industrial Actuators

Comparison with data of industrial actuator can provide very relevant information.
Firstly, it will illustrate whether the actuator behaves similarly to other actuators of
the same class. On the other hand, the designer will be able to decide if the actuator
performs better than other competing actuators or whether the same performance is
achieved but at lower cost, weight or volume.
62 2 Actuator Design Analysis

2.9.3 Simulation

The simulation of the system under study can provide relevant information in or-
der to validate the conclusions extracted in the previous steps. Moreover, computer
aided engineering (CAE) tools allow the designer to test the proposed actuators
without the need for building a real prototype. Different particular cases can be
studied by using the obtained optimal geometries and dimensions. In such cases,
not only the relevant output quantities can be validated, assuring that the obtained
analytic expressions hold for a particular case, but also the assumed limiting quan-
tities can be checked and it can be observed which are the most critical areas of the
actuator.
Finite element analysis (FEA) [19, 7, 16] employs the nite element method
(FEM) to deal with a broad range of engineering applications. There are a num-
ber of nite element simulation packages like COMSOL , ALGOR , ANSYS ,
NASTRAN , etc. Most of such packages are based on the virtual work principle
or the minimum total potential energy principle. Finite Element Analysis was rstly
proposed in 1943 by Richard Courant who established the mathematical basis. In the
1950s and 1960s it started to be employed in engineering applications, starting with
mechanical engineering. Nowadays, nite element analysis is a fundamental tool
for systems simulation, since it has changed deeply the way engineers approach the
design of their applications.
In order to simulate using nite element analysis a certain actuator along with the
system and structure where it is integrated, the following steps have to be followed:
1. Denition of all the constants involved in the simulation, including the so-called
universal constants but also the material properties assumed. Most packages
include material library where the different material properties are already in-
cluded.
2. Denition of the assumed symmetries in order to simplify the analysis. The
higher the complexity of the system under study, the more important to consider
symmetries, studying a single portion and extract conclusions for the overall
system.
3. Drawing (or import) of the system under analysis. In order to obtain relevant
results it is extremely important to have an accurate drawing of the studied
actuator. The drawings can be also imported from common CAD packages.
4. Specication of the boundary conditions. Typically the surrounding of the sys-
tem has to be included in the drawing, since it also plays an important role in the
physical system. Air, for example has a certain magnetic permeability, so that
the magnetic ux can ow through it. The surrounding system limits have to
be clearly dened stating their boundary conditions. For example, in an system
analyzed from the electromagnetism point of view, it is common to establish
electrical or magnetic isolation as boundary conditions.
5. Specication of the physics laws to be considered in the analysis. Each system
part is ruled by different physics laws. It has to be clearly established what is
the physical law ruling each system block.
2.9 Validation 63

6. Denition of the nite element mesh, considering different types of mesh ac-
cording to the importance of the region. Most critical regions have to have the
nest a mesh.
7. Problem solving. Once all the previous steps have been properly addressed, the
system can be solved, by rstly tuning the solver appropriately, specifying what
kind of analysis is to be done.
8. Post-processing. The obtain results can be plotted so that the desired results can
be easily graphically seen.
All these steps are illustrated with a simple system containing a permanent magnet
and a levitating ball:
1. Constant denition. The following constants have been dened: permeability
of vacuum 0 = 4107 NA2 , permanent magnet magnetization M pm = 750
103 A/m and the magnetic relative permeability r = 15000.
2. Denition of the assumed symmetries. In this examples no symmetries are con-
sidered, since it is a very simple system.
3. Drawing. The drawing can be seen in Figure 2.23. The permanent magnet is a
block of 40 mm 40 mm 50 mm, the ball is a sphere of radius 20 mm whose
closest point to the permanent magnet is placed at 5 mm.
4. Specication of the boundary conditions. The air surrounding the permanent
magnet and the ball includes a large region, whose boundary conditions estab-
lish magnetic isolation as shown in Figure 2.23.
5. Specication of the physics laws to be considered in the analysis. The air and
the ball are ruled by
0 r Vm = 0 (2.60)
where Vm is the so-called magnetic potential. The so-called constitutive law
yields B = 0 r H with the difference that while the ball has a high r = 15000,
the air has r = 1. The permanent magnet is governed by

(0 Vm 0 M pm ) = 0 (2.61)

with the constitutive law B = 0 H + 0 M pm . The stress and force in the ball are
the relevant output quantity, therefore it has to be specied that such electro-
magnetic stress and force is to be computed.
6. Denition of the nite element mesh. As illustrated in Figure 2.24, the system
mesh is performed, considering a ner mesh for the permanent magnet and the
ball.
7. Problem solving. The system is solved using the stationary linear solver.
8. Post-processing. The obtained results are graphically shown as illustrated in
Figure 2.25 using streamlines for the magnetic ux and boundary colors for
the ball electromagnetic stresses, Figure 2.26 using boundary colors and arrows
related to the Maxwell tensor stresses, Figure 2.27 where slices are used to plot
the magnetic ux density and Figure 2.28 where the a number of isosurfaces
show the points having the same magnetic ux density.
64 2 Actuator Design Analysis

Fig. 2.23 Typical screen of a nite analysis software (COMSOL by COMSOL AB) of the ex-
ample of the permanent magnet and the levitating ball
2.9 Validation 65

Fig. 2.24 Example element

mesh using COMSOL by
COMSOL AB. It can be noted
that the permanent magnet
and the ball are ner meshed
that the surrounding air

Example 2.15. Analyze the force-position characteristic of ball levitating due to the
attraction of a permanent magnet.
The force expression of a permanent magnet attracting a levitating ball cannot be
easily expressed analytically. A simplied expression can be used
1
FM = (2.62)
Ni=0 ai zi
In order to determine the optimum degree N and the coefcients ai different
simulations have been developed for different distances z. The obtained results are
shown in Table 2.3 and Figure 2.29.

Table 2.3 Force-position data obtained using FEA

z [mm] 0 0.5 1 1.5 2 3 5 7 10 15 20 30
F [N] 36.2 31.71 27.66 24.25 21.45 16.46 10.1 6.2 3.13 1.12 0.45 0.1

Using standard regression techniques, FEA results can be analyzed with different
degree polynomials. It can be observed in Figure 2.30 that the best matching is
obtained for N = 5. The obtained force expression yields
1
FM (z) = (2.63)
c0 + c1 z + c2 z2 + c3 z3 + c4 z4 + c5 z5
66 2 Actuator Design Analysis

Fig. 2.25 Post-processing of the solved examples. The streamlines show the magnetic ux ow
2.9 Validation 67

Fig. 2.26 Post-processing of the solved examples. The ball colors and arrows show the different
Maxwell tensor stresses in the ball
68 2 Actuator Design Analysis

Fig. 2.27 Post-processing of the solved examples. The slices show the different values of magnetic
ux density
2.9 Validation 69

Fig. 2.28 Post-processing of the solved examples. The isosurface show the different surfaces hav-
ing the same magnetic ux density
70 2 Actuator Design Analysis

40

z=0mm F=36.2N
35

30

25
Force [N]

20

15

10

0
0 5 10 15 20 25 30
Distance z [mm]

Fig. 2.29 Obtained FEA results for the permanent magnet and ball system of Example 2.15

with c0 = 1.47 107 , c1 = 5.14 106 , c2 = 3.35 105 , c3 = 0.0013, c4 =

0.0066 and c5 = 0.028.

FEA data
Polynomial degree 3
Polynomial degree 4
1
Polynomial degree 5
10
Force [N]

0
10

1
10
0 5 10 15 20 25 30
Distance z [mm]

Fig. 2.30 Comparison of the different polynomials proposed to model the force-displacement
curve of Example 2.15
2.10 Considerations on Actuators Dynamics 71

2.10 Considerations on Actuators Dynamics

2.10.1 Dynamical Analysis

If the dynamical quantities are considered, it is necessary to consider not only the
actuator but the load attached to it. A typical linear load can be characterized by an
elastic constant kL , damping coefcient bL and mass mL . Considering the actuator
force Fa (t), the dynamics differential equation yields:

Fa (t) = mL x(t) + bL x(t) + kL x(t) (2.64)

where x(t) is the actuator displacement.

The transfer function between the displacement and the force can be expressed
as
X(s) 1
= (2.65)
Fa (s) mL s2 + bL s + kL
where s is the Laplace operator and X(s) and Fa (s) are the Laplace transforms of
x(t) and Fa (t).
Analyzing the transfer system poles according to linear systems theory [8], the
following system poles are obtained:

bL b2L 4mL kL
s12 = (2.66)
2mL
where it can be clearly noted that the system dynamics will importantly depend
on
b2L 4mL kL (2.67)
The natural frequency 0 0 and the damping ratio may be dened as

kL
0 = (2.68)
mL
bL
= (2.69)
2 kL mL
Substituting (2.68) and (2.69) in (2.66) the poles yield
  
s12 = 0 2 1 (2.70)

Depending on the value of different system behaviors can be considered:

= 1 The system is critically damped. There is double pole at 0 , assuring
stability for 0 > 0. The homogenous solution of the system dynamic yields

x(t) = (k1 + k2t) e0 t (2.71)

72 2 Actuator Design Analysis

where k1 and k2 depend on the initial conditions x(0) and x(0).

> 1: the system is over-damped. Both poles are real and the system will be
stable as long as both poles have negative real part. Hence, it will always be
stable since 
2 1 < 0 (2.72)
for any > 1. The homogenous solution of the system dynamic yields
   
0 2 1 t 0 + 2 1 t
x(t) = k1 e + k2 e (2.73)

where k1 and k2 depend on the initial conditions x(0) and x(0).

< 1 The system is under-damped. The poles yield

0 j0 1 2 (2.74)

where the pole negative real part for > 0 guarantees stability. The homogenous
solution of the system dynamic yields
      
x(t) = e 0 t k1 cos 0 1 2t + k2 sin 0 1 2t (2.75)

where k1 and k2 depend on the initial conditions x(0) and x(0).

Real part of system poles

0
0 (+(21)1/2)
50

100

150

200

250

300
0 ((21)1/2)
350

0
0.5 100
1 80
60
1.5 40
20
2 0
0

Fig. 2.31 Real part of the poles of (2.70)

The system poles for the different poles can be observed in Figures 2.31 and 2.32.
The step response of under, over and critical damped systems is illustrated in Fig-
ure 2.33 for different values of and 0 = 10 rad/s. It can be noted that there is
only one case for = 0 where the displacement oscillates permanently since it has
2.10 Considerations on Actuators Dynamics 73

Imaginary part of system poles

100 ((21)1/2)
0

50

50

100
100 0 (+(21)1/2)
80 2
60 1.5
40 1
20 0.5
0 0
0

Fig. 2.32 Imaginary part of the poles of (2.70)

both poles with real part equal to zero. The frequencial response is illustrated in the
bode diagram of Figure 2.34, where it can be noted that the resonance frequency of
the system corresponds to the natural frequency 0 , and it is important for poorly
damped systems.

2.10.2 Control System

In previous Section 2.10, the dynamics of the load have been described. When the
whole system considering also the actuator is to be analyzed, it is important to in-
clude the control system. The control system can be either an open loop control
system or a closed loop control system:
In open loop, the actuator input quantities are established arbitrarily without feed-
back and therefore without verifying that the control output quantity is actually
achieving the desired value. Most open loop applications are undertaken in open
loop because no sensor for the output quantity is needed, with the cost reduction
involved. The main drawback is the lack of precision. A typical open loop system
is sketched in Figure 2.35.
Closed loop system measure or estimate the output quantity and feed it back to
the control system in order to eliminate the so-called controller error. A typical
closed loop system is sketched in Figure 2.36.
The block diagrams of Figures 2.35 and 2.36 illustrate typical open loop and
closed loop control systems. The main elements are:
74 2 Actuator Design Analysis

=0 underdamped
0.02 =0.1 underdamped
=0.5 underdamped
=1 criticallydamped
=1.5 overdamped
0.015
=5 overdamped
x [m]

0.01

0.005

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

Time [s]

-7
x 10

1 ]=0 under-damped
]=0.1 under-damped
]=0.5 under-damped
0.8
]=1 critically-damped
]=1.5 over-damped
0.6 ]=5 over-damped

0.4

0.2
v [m/s]

-0.2

-0.4

-0.6

-0.8

-1

0 0.005 0.01 0.015 0.02

x [m]

Fig. 2.33 Dynamic step response of the system with different and 0 = 10 rad/s

The power grid (1) provides the energy needed to drive the actuator (3). It can
be by means of electricity, uid power, heat, etc.
The power conversion (2) unit takes the output of the control system (5) and
drives the actuator. Typical examples of converters are electrical inverters or hy-
draulic proportional valves.
The actuator (3) transforms the energy supplied into the mechanical energy ap-
plied to the load (4).
2.10 Considerations on Actuators Dynamics 75

0
10 =0 underdamped
Magnitude =0.1 underdamped
=0.5 underdamped
=1 criticallydamped
5
10 =1.5 overdamped
=5 overdamped

2 1 0 1 2 3 4
10 10 10 10 10 10 10
[rad/s]

0 =0 underdamped
=0.1 underdamped
50
=0.5 underdamped
Phase

100 =1 criticallydamped
=1.5 overdamped
150 =5 overdamped

200 2 1 0 1 2 3 4
10 10 10 10 10 10 10
[rad/s]

Fig. 2.34 Bode plot of the system for different and 0 = 10 rad/s

Fig. 2.35 Typical open loop control block diagram of a mechatronic system

The controller (5) takes the set point values referenced by the operator (6) and
applies the obtained control effort output to the power conversion unit (2).
It can be noted that the energy can ow in both directions in many applications.
Although actuators are designed to behave transforming a certain energy source into
mechanical energy in some occasions they have to operate as generators and trans-
form mechanical energy in the source energy. Actuators featuring such bidirectional
energy ow allow to brake or return to the original position without the need for ex-
76 2 Actuator Design Analysis

Fig. 2.36 Typical closed loop control block diagram of a mechatronic system

ternals brakes. Furthermore, they are more efcient since part of the energy may be
returned to the power grid. A typical example of these efcient reversible machines
are electrical machines.
The object of the present book is not to delve into control aspects of dynamical
systems, and hence some assumptions can be done in order to simplify the analysis:
1. The controller can be executed in continuous time. It is important to remark that
it is not the case when digital discrete time systems are concerned. To delve into
such discrete time systems [10] can be consulted.
2. The analyzed systems are considered linear[8], i.e. the state variables x(t) of the
system can be expressed as functions of the inputs u(t) as

x(t) = Ax(t) + Bu(t) (2.76)

y(t) = Cx(t) + Du(t) (2.77)

where y(t) is the system output and A,B,C and D are the system matrixes. If
at least one of such matrixes vary with time, the system is considered Linear
Time-Varying (LTV) or Linear Parameter-Varying (LPV), otherwise if the four
matrixes do not vary with time, a Linear Time-Invariant (LTI) system is consid-
ered. If the system is not linear, it can be linearized used standard linear systems
tools [8].
Nonlinear systems are carefully studied in [9]. Their state variables can be ex-
pressed in the general form:

x(t) = (x(t), u(t)) (2.78)

2.10 Considerations on Actuators Dynamics 77

where it can be noted that the function represent a number of different non-
linearities. Depending on the function , the appropriate analysis procedure can
be used [9].
3. No perturbations are considered.
4. The control effort (controller output to be applied by the actuator) is unbounded
or its bounds are not overcome during the analyzed time. To consider systems
not satisfying such assumption, see [3].
Although control theory is an extremely broad engineering discipline, and there
exist numerous different controllers to be applied in active systems, it is common
dene the error (t) = x (t) x(t) as the difference between the reference value
x (t) and the measured value x(t) and the output control effort as function of such
an error as   
y = f (t), (t), (t)dt (2.79)

According to the previous assumptions, the system (2.64) can be studied

  
f (t), (t), (t)dt = mL x(t) + bL x(t) + kL x(t) (2.80)

Although there are a number of available families of controllers, the simpler and
more extensively used controllers are the so-called PID controllers. Such controllers
compute the output as
   
f (t), (t), (t)dt = KP (t) + KI (t)dt + KD (t) (2.81)

where KP , KI and KD are controller constant, which are to be designed analyzing

carefully the system under study.

Example 2.16. Consider a control law

  
f (t), (t), (t)dt = k p (t) (2.82)

with a reference value x (t) = 0. Discuss how the controller can modify the response
of the system.
Substituting the control law in the system (2.80)

k p x(t) = mL x(t) + bL x(t) + kL x(t) (2.83)

It is straightforward to note that the new system dynamics will be the same as the
unactuated system with a modied elastic constant

kL = kL + k p (2.84)

The new system dynamics will be characterized by:

78 2 Actuator Design Analysis

kL + k p
0 = (2.85)
mL

bL
=  (2.86)
2 (kL + k p ) mL
allowing to freely choose the system response by choosing the adequate constant
k p . Note that the while the sign of k p can be chosen the system will be stable as long
as k p > kL .

References

1. Buckingham E (1914) On physically similar systems: Illustrations of the use of dimensional

equations. Phys Rev 4:345376
2. Holman J (1980) Thermodynamics, 3rd edn. McGraw-Hill, Kogakusha
3. Hu T, Lin Z (2001) Control Systems with Actuator Saturation: Analysis and Design.
Birkhauser
4. Huber J (1998) Ferroelectrics: Models and applications. PhD thesis, University of Cambridge
5. Huber JE, Fleck NA, Ashby MF (1997) The selection of mechanical actuators based on per-
formance indices. Proc R Soc Lond A 453:21852205
6. Jack A, Mecrow B (1990) Methods for magnetically nonlinear problems involving signif-
icant hysteresis and eddy currents. IEEE Transactions on Magnetics 26(2):424429, DOI
10.1109/20.106344
7. Jin J (1993) The nite element method in electromagnetics. Wiley New York
8. Kailath T (1980) Linear systems. Prentice-Hall
9. Khalil H (2002) Nonlinear systems. Prentice Hall Upper Saddle River, NJ
10. Kuo B (1992) Digital Control Systems. Oxford university press
11. Kuribayashi K (1993) Criteria for the evaluation of new actuators as energy converters. Ad-
vanced robotics 7(4):289307
12. Kyriakides E, Heydt G, Vittal V (2005) Online parameter estimation of round rotor syn-
chronous generators including magnetic saturation. IEEE Transactions on Energy Conversion
20(3):529537, DOI 10.1109/TEC.2005.847951
13. Levi E (1997) Modelling of magnetic saturation in smooth air-gap synchronous machines.
IEEE Transactions on Energy Conversion 12(2):151156, DOI 10.1109/60.629697
14. Nandi S (2004) A detailed model of induction machines with saturation extendable for fault
analysis. IEEE Transactions on Industry Applications 40(5):13021309
15. Novotnak R, Chiasson J, Bodson M (1999) High-performance motion control of an induction
motor with magneticsaturation. IEEE Transactions on Control Systems Technology 7(3):315
327
16. Reddy J (1993) An introduction to the nite element method. McGraw-Hill Singapore
17. Sullivan C, Kao C, Acker B, Sanders S (1996) Control systems for induction machines with
magnetic saturation. IEEE Transactions on Control Industrial Electronics 43(1):142152
18. Taylor E (1974) Dimensional analysis for engineers. Clarendon Press Oxford
19. Zienkiewics O, Taylor R, Nithiarasu P, Zhu J (2005) The Finite Element Method.
Elsevier/Butterworth-Heinemann
http://www.springer.com/978-1-84882-613-7