Sie sind auf Seite 1von 133






(Disegno di meccanismi miniaturizzati articolati attraverso cerniere flessionali)
settore scientifico disciplinare: ING-IND/15




My special gratitude goes to prof. Danilo Cambiaghi and to Mr. Valerio Villa who have trusted me from the
beginning of my PhD.
I address my special thanks to all the members of the Mechanical Draw Group of the University of Brescia
for the advice, the help and the friendship found: thanks to Mr. Gabriele Baronio, Mr. Stefano Uberti,
Mr. Riccardo Metraglia and prof. Emilio Chirone.
I would also like to thank all the members of the Industrial and Mechanical Department of the University of
Brescia who helped me in this three years: specially prof. Laura Depero and prof. Elza Bontempi, who have
given me the possibility to approach, for the first time, the compliant mechanisms and Mr. Matteo Lancini,
who supported me during the setting of the Bydlo device.
It is a pleasure to address my special thanks to Mr. Simon Henein who took me under his wing in
Switzerland and who shared with me his knowledge about flexure mechanisms.
Many thanks also to the CSEM that has given me the possibility to work for nine months in Neuchtel:
special thanks to Mr. Jean-Marc Breguet, Mr. Peter Spanoudakis, Mr. Ivar Kjelberg, Mr. Philippe Schwab,
Mr. Laurent Giriens, Mr. Serge Droz, Mr. Yvon Welte, Mr. Jean-Michel Mayor, Mr. Mario El-Khoury, and to
all the colleagues of the Systems Engineering Division.
In the end, many thanks to Mrs. Jodi Villa who has provided me for my terrible English.

Thanks. Merci. Grazie.

Luca Dassa

After a fast review of the state of the art (chapter 1) as regards to the so-called compliant mechanisms
with flexure hinges, chapter 2 tries to clarify the terminology, introducing at the same time Heneins
approach to the mechanism based on flexure articulations.
Chapters 3 and 4 show the first designed devices, my first steps in this research field: a miniaturized
machine for traction/compression and the translation of the universal joint kinematics in flexure structure.
In particular the machine for traction/compression was manufactured and the first tests are reported here.
The naiveties due to the basic knowledge in flexure field are also presented and discussed.
The mechanism for space application (chapter 5), the 2-DOFs nanoconverter (chapter 6) and the MEMS for
optical applications (chapter 7) are three projects designed during or after my internship at the CSEM
(Centre Suisse dlectronique et microtechnique Neuchtel - Switzerland): a different approach to the
flexure mechanisms is presented. Above all, the study of the MEMS device is deepened until the test phase
and represents a very interesting attempt to transfer the rules in designing flexure structures from the
traditional scale to the MEMS scale.
The last chapter touches briefly on the design of flexure structures in plastics.





Flexure devices based on piezo actuators


Devices for very high precision placement


Measure sensors


Mechanisms for space applications


Parallel kinematics robots




Devices manufactured in plastics.




Howells approach



Lobontius approach



Heneins approach








The Bydlo linear bearing



Partial models



The traction/compression machine



Results of the preliminary tests



Analysis with Heneins point of view








The design of a compliant universal joint



3UPU parallel kinematics robot



Dimensioning of the compliant joint for the 3 DOFs robot



Analysis with Heneins point of view.



Use of the flexible universal joint





Choice of the actuator and the sensor



Preliminary asymmetric design



Symmetric design



PAAM dimensioning and material choice



Solution to decrease the effects of the driving rod



Different solution for the pivot





The original nanoconverter



The 2 linear DOFs nanoconverter



The 3 linear DOFs nanoconverter






Errore. Il segnalibro non definito.

Study based on existing generations
System description
Comb actuators
Optical deformable area

Errore. Il segnalibro non definito.

Errore. Il segnalibro non definito.
Errore. Il segnalibro non definito.
Errore. Il segnalibro non definito.

Comb guiding system

Errore. Il segnalibro non definito.

Errore. Il segnalibro non definito.
Errore. Il segnalibro non definito.
Errore. Il segnalibro non definito.

Simulations of the full structure for previous generations

Errore. Il segnalibro non definito.

Errore. Il segnalibro non definito.
Errore. Il segnalibro non definito.



Conclusions and input for the design of a new generation

Errore. Il segnalibro non definito.

Errore. Il segnalibro non definito.

New generation analysis (vast04)
Errore. Il segnalibro non definito.
Comb actuator
Errore. Il segnalibro non definito.
Design of the optical area
Errore. Il segnalibro non definito.
Extra-lateral anchors
Errore. Il segnalibro non definito.
Comb guiding mechanism design
Errore. Il segnalibro non definito.
Connection elements between optical area and combs systemErrore. Il segnalibro non
New generation devices
Deformable device with big optical area (1x1 mm2)
Deformable device with small optical area (0.5x0.5 mm2)

Errore. Il segnalibro non definito.

Errore. Il segnalibro non definito.
Errore. Il segnalibro non definito.

Tests on the implemented solutions
A1 device
B1 device
D1 device
D2 device
F1 device
K2 device
Blade thickness analysis
Results discussion

Errore. Il segnalibro non definito.

Errore. Il segnalibro non definito.
Errore. Il segnalibro non definito.
Errore. Il segnalibro non definito.
Errore. Il segnalibro non definito.
Errore. Il segnalibro non definito.
Errore. Il segnalibro non definito.
Errore. Il segnalibro non definito.
Errore. Il segnalibro non definito.


Conclusions and future developments


Errore. Il segnalibro non definito.



Laboratory pincer



Child highchair






The research on compliant mechanisms with flexure hinges was started from 2 projects developed by prof.
Danilo Cambiaghi and eng. Valerio Villa at the Mechanical and Industrial Engineering Department
University of Brescia.
The Cambiaghis device [1], shown in Fig. 1 and in Fig. 2, was related to space application. It was necessary
to place a pincer under vacuum on a satellite. Very little material was available because of the
environmental condition and the traditional mechanical design, such as ball bearings, was not advised due
to the lubrication problems. The hard work conditions pushed to find new solutions for the device: the
problem was smartly solved using flexible elements manufactured in stainless steel. The kinematics is very
simple but efficient: a limit was introduced into the device to shield the flexible blade and to avoid an
exaggerated deformation. In Cambiaghis design, the kinematics is very simple: a possible schema with rigid
link is shown in Fig. 2. Every flexible element is replaced with an ideal pivot coupled with a rotational
spring. At the same time also the stiffness of the hinge was exploited to guarantee the shell hold during
stand-by and to allow the return of the pincer in the closed position after opening. The Eureca satellite flew
and came back on Earth; the mechanism was open and it was possible to verify the correct functioning of
the pincers.

Fig. 1 the Eureca


Fig. 2: the Eureca pincer


Fig. 3: the compliant

drafting compass

Fig. 4: the compliant drafting

compass (scheme)

The Villas drafting compass [1] is made from a different material, a moulded polyamide reinforced with
glass fibres (Fig. 3). The aim of this design is completely different from the previous device because this
application is a mass device produced in thousands. The kinematics is more complex than the previous
device, as shown in Fig. 4. In this case, it should also be possible to replace the flexible elements with ideal
joints coupled with springs but because every hinge is loaded beyond the elastic limit, the low resulting
stiffness can be neglected. To verify the reliability, the drafting compass was tested for one million cycles
without failure. In this case the choice to use a compliant device allows reducing the assembly costs,
producing in one time the complete mechanism.
These two projects are very interesting for the opened perspectives: the new approached research field, in
the case of the Cambiaghis project about 25 years ago, is the topic of this PhD thesis.
1 / 126

From these applications it is possible to excerpt some preliminary considerations.

The compliant mechanisms present very interesting features, to be used in very different fields. The specific
characteristics can be useful for high performances, such as space applications, and for low cost devices,
such as the drafting compass. For sure it will be necessary to use different characteristics and different
material for each field: in difficult environment it is possible to exploit the absence of mechanical backlash
and lubrication, in low cost devices it is possible to exploit the possibility to reduce assembly problems and
Another basilar consideration is that the designer always tries to find an equivalence between compliant
mechanism and a model with rigid bodies and ideal joint coupled with springs: different researchers have
tried to find a way to perform it simply. Later on different approaches will be compared.

The aim of my PhD thesis is to deepen the knowledge about the compliant mechanisms and above all about
the compliant mechanisms with flexure hinges. The review of the literature shows that the division
between compliant mechanisms and compliant mechanisms with flexure hinges is not so clear: a guideline
able to solve the question will be founded in the publication of Mr. Simon Henein.
My knowledge about the flexible mechanisms has greatly increased thanks to the 9 months spent at the
CSEM (Centre Suisse dlectronique et microtechnique Neuchtel - Switzerland) during the last year of my
PhD: the Systems Engineering Division of the CSEM, where I worked, is focused on the design of flexure
based mechanisms. Moreover at the CSEM I had the possibility to work directly with Mr. Simon Henein, one
of the most important researchers in this field.
The theoretical fundamentals of Heneins approach will be explained in the first part of the thesis and then
deepened in the whole thesis. In agreement with the philosophy of the Mechanical Draw Group where I
have spent most of my PhD, some different applications of flexure based mechanisms will be presented,
explaining for each one the design process, from the idea to the definition of the model and to the device
drawing. This thesis is mainly composed of a wide collection of exempla of devices with flexures: where the
applications were developed before meeting Mr. Henein, I will try to review the project with a point of view
closest as possible to Heneins approach.

2 / 126

1 State of the art

In agreement with the scientific community, the mechanisms with flexure hinges are a particular class of
the complaints mechanisms: the compliant mechanisms are mechanisms where the mobility is allowed by
the deflection of flexible members and the deformation is diffused into the mechanism. In flexure based
compliant mechanisms the flexibility is concentrated in specific areas, named flexure hinges, that is the
flexible members are very small with respect to the whole mechanism. In the literature, the terminology
flexure hinge is not so clear: it is possible to find terms such as flexural hinges, flexural pivots, living
hinges, notch hinges, with a vague distinction. In the Fig. 5 and in Fig. 6 there are two exempla:
intuitively, seeing the figures the difference between compliant mechanism and flexure based mechanism
should be clear. Both mechanisms are based on the deltoid Q-joint (see [3], at page 185), where each rigid
segment in the quadrilateral is made adjacent to a segment of equal length. In the medical forceps (see Fig.
7), the deltoid Q-joint is achieved using flexure hinges; in the compliant scissors (see [4]), it is more difficult
to find the pivots rotation axes (see Fig. 8): the motion is the result of the deformation of a big part of the
mechanisms. It is necessary to observe that the distinction between compliant mechanisms and flexure
based mechanisms is very clear in this case, because the exempla show two opposite conditions.

Fig. 5: medical forceps

Fig. 6: compliant scissors

Fig. 7: medical forceps (scheme)

Fig. 8: compliant scissors (scheme)

In general the division is not so clear. Very often the flexible members of the compliant mechanisms
deform beyond the limit of small displacements, while in general for the flexure hinges the situation is the
opposite. Moreover the metallic compliant mechanisms are used under the elastic limit, while the plastic
materials can be exploited beyond this limit. With these considerations, does the device [5] in Fig. 9 have to
be considered compliant mechanism or should it be considered mechanism with flexure hinges? In fact,
it is quite difficult to find some flexure hinges, with respect to the previous definition, since the flexible
elements are quite long with respect to the size of the whole device. So, it should be better to use the term
compliant mechanism, even if the hypothesis of small displacement is valid. But is it possible to amplify
the field of the so-called flexure hinges to contain this device as well? After a fast review of some
application where the flexure hinges and the compliant mechanisms are employed, there will be an
attempt to answer to this question.
3 / 126

Fig. 9: butterfly pivot

1.1 Flexure devices based on piezo actuators

Historically, one of the most important fields for the flexure mechanisms is the design of piezo driven
devices. In fact one of the most performing actuator in micrometric scale is the piezo actuator. Since it is
impossible to couple it with traditional joints, from the beginning it was necessary to find a guiding
mechanism for piezo driven devices and the solution was found in the mechanisms based on flexures. At
the same time because the motion range of a piezo actuator is very small, there was the necessity to
amplify the motion: this requirement can be achieved using flexure mechanisms as well. Today there are
some companies for which the piezo actuators and guiding systems are the core business, Physik
Instrumente above all [6].
An exemplum of the PI products is the P-780 piezo-driven, flexure-guided stage (see Fig. 10). Providing a
positioning and scanning range of up to 80 m with settling times of only a few milliseconds, the P-780 is
designed for applications with loads up to 100 g. Closed-loop and open-loop versions are available. In Table
1 there are the most important features of this device. Neglecting a complete kinematics analysis, a fast
glance to the picture reveals the concept: a linear stage based on flexural elements is driven by a lever,
manufactured with flexure hinges as well. The piezo, located near the stage, acts directly on the lever.

Fig. 10: the P-780 stage

4 / 126

Table 1: P-780 main features

80 m 20%
80 m
20 nm
1.5 N/ m 20%
50/5 N

Open loop travel @ 0 to 100 V

Closed-loop travel
Integrated feedback sensor
Closed-loop / open loop resolution [-/1]
Closed-loop linearity
Full-range repeatability
Push/pull force capacity in operating
Max () normal load
Lateral force limit
Unloaded resonant frequency
Resonant frequency with 100 g load
Weight with cables
Overall dimensions

10 N
10 N
1000 Hz 20%
600 20%
170 g 5%
12,5x40x50 mm3

It is possible to find other commercial actuators, such as the DSM FPA-500E [7]. The mechanics (see Fig. 11)
is very simple: the flexure stage has the purpose to amplify the displacements; the steel wire preloads the
two piezos. In Table 2 there are the main specifications of the device. The amplification of the displacement
is based on the scheme shown in Fig. 12, where the two ends of the rigid link AB have to move along two
orthogonal directions: this linkage is a four-bars linkage where the two ground hinges are replaced by
prismatic joints (= hinges with rotation axis at the infinity).

Fig. 12: the DSM FPA-500E (scheme)

Fig. 11: the DSM FPA-500E

Table 2: the DSM FPA-500E main features


500 m 10%


0.22 N/m 10%

Unloaded resonant frequency:

440 Hz 5% (fixed-free)

Blocked force estimate:

110 N


19 x 52 x 10 mm

5 / 126

1.2 Devices for very high precision placement

The requirements in terms of high precision displacement has provided the manufacturing of very
interesting mechanisms in bigger scale than piezo systems. One of the most important company using
flexure devices for precision mechanism is the CSEM, Centre Suisse dlectronique et de microtechnique.
The excellence of the CSEM in the field is proved by the numerous publications signed by its members. The
CSEM is the actual workplace of Mr. Simon Henein, already introduced such as one of the most important
researcher in the flexure field. In Fig. 13, there is the extremely high-resolution tip-tilt-piston (3 DOFs)
mirror mechanism for the VLT-NAOS field selector [8]: the mechanism provides a mechanical angle
amplitude of 6 with a resolution and mechanical stability of 0.42 arcsec rms over 20 minutes. The
envelope is a 125 x 180 mm cylinder. The configuration is based on three electromagnetic actuators 120
apart with the mobile magnets mounted on flexure guides. All kinematic joints consist of flexure elements
so that the mechanism is essentially frictionless. In Fig. 15 it is possible to see the schema of the device. A
central membrane in CuBe (see Fig. 14) allows the tip-tilt movement around its centre of rotation,
providing radial rigidity and stability.
output link

Fig. 14: the NAOS membrane

fixed link


Fig. 13: the NAOS device

Fig. 15: the NAOS scheme

1.3 Measure sensors

Fig. 16: flexure component (Mettler-Toldo balance)

6 / 126

One of the most famous and beautiful devices for measurement purposes is the mechanism used in the
Mettler-Toldo balance [9] (Fig. 16), where the force from the mass is geared down by two levers before
the compensation, applied with electromagnetic actuators. The balance based on this device is currently in
production. With this design it is possible to increase the balance accuracy, without adding the complexity
typical of the manufacturing of the two levers with standard joints.

1.4 Mechanisms for space applications

The flexure based mechanisms are often used in spatial devices because of some very performing features
such as the absence of the lubrication and the high reliability.
Between the different publications, the piezoelectric XYZ stage, used in the Atomic Force Microscope
(AFM), placed on the Rosetta space probe [10] is shown in Fig. 17 and in Fig. 18. This is an hybrid
parallel/serial mechanism: the 2 DOFs x and y are provided by a parallel stage, moved in each direction by
four piezos with amplifier; on the mobile link there is the piston stage, a pre-charged simple piezo actuator.
The functional performances achieved are 100 m of stroke in the X and Y directions, 8 m in the Z
direction, with minimized parasitic rotations: z < 240 rad, and x, y < 20 rad.

Fig. 18: : stage for Rosetta space probe (3D model)

Fig. 17: stage for Rosetta space probe (photo)

1.5 Parallel kinematics robots

It was clear from the beginning that the disadvantages of the parallel configurations were not so dangerous
in the robots with flexible structures. In fact, since the motion is limited by the flexure joints, in a flexure
based parallel robot, it is almost impossible to approach to the singular positions and it is impossible to
have interferences between different links. In the opposite way, it is possible to exploit the advantages
such as high stiffness, high dynamics, actuators linked to the ground, high first frequency.
Between different publications, special mention deserves to the Orion robot (see Fig. 19), developed in the
LSRO in Lausanne [11]. It is a 3 DOFs parallel kinematics robot based on 3 identical kinematics chains which
produces movements in x, y and Z. The model has been designed to constitute the left hand of a machine
tool that requires orienting the workpiece in a very precise manner and with high rotation amplitude: the
model has been designed with joints achieving 8 leading to an output angle of 15 on both rotation axes.
In Z direction the displacement is not limited by the flexures but only by the guiding device of the linear

7 / 126

Fig. 19. Orion robot

In Fig. 20 and in Fig. 21 it is possible to see the nano-stage mounted on the machine AGIETRON Micro-Nano
for micro wire EDM machining [12]. Presented in 2005 this machine was the most performing in the world.
In this machine instead of the axis C (vertical axis), there is mounted a parallel stage X, Y, Z, where each axis
is moved by a Voice Coil actuator, and the measurement is performed by 3 optical rules. The most
important features are the following:

displacements (X x Y x Z): 6 x 6 x 6 mm

Precision of positioning: 0.1 m

Resolution: 0.02 m

Min roughness: Ra 0.05 m

Fig. 20: AGIE Agietron Micro&Nano

Fig. 21: AGIE Agietron Micro&Nano (detail)

8 / 126

1.6 MEMS
The compliant mechanisms are already diffused in MEMS field but it is not usual to find flexure hinges
based MEMS. One of the main reasons is that the displacements required to the device are big with respect
to the overall dimensions of the device. Some interesting devices are presented in Overstolz PhD thesis
[13]. This PhD thesis presents the realization of micro electro-mechanical systems for optical applications
requiring very precise actuation. Two devices have been realized following two different concepts for
vertical comb drive actuators. One device provides a tilting platform of 22mm2 (see Fig. 22 and Fig. 24)
with to the possibility to set a tilt angle in a range of 3.5. The other device provides a triangular mirror of
1.7mm edge length which is suspended by three beams (see Fig. 23 and Fig. 25). The mirror, actuated by
three sets of asymmetric vertical comb actuators, can perform either a pure vertical (piston) displacement
of up to 18 m, or a tip-tilt of up to 2.2 mrad. The Overstolz thesis performs an analysis of the suspension
beams in every case, with analytical formula and with FE software.
In the first device, two elementary articulation allow the displacement with their torsion; in the second
device, the suspension (similar to the membrane shown in Fig. 14 and in Fig. 15) is based on the flexion of
three rods (the beams have an aspect ratio of beam height vs. beam width of 1:2 - 1:3).

Fig. 22: 1-DOF mirror

Fig. 24: 1-DOF mirror (photo)

Fig. 23: 3-DOFs mirror

Fig. 25: 3-DOFs mirror (photo)

1.7 Devices manufactured in plastics.

The use of the hinges in plastic material is quite spread in packaging industry: it is sufficient to think to the
numerous kind of food packaged with monolithic box, to be wasted after eating the content. But the hinges
are used also in boxes for multiple use, such as the boxes shown in Fig. 26, Fig. 27 and Fig. 28, sold in Italy
9 / 126

by Tupper Ware S.p.A. [14]. In this case the device is very simple, usually composed of only one flexure
hinge, manufactured by moulding process. In this case it is important to have devices able to stand a
sufficient cycle number.

Fig. 26: compliant box cover (1)

Fig. 27: compliant box cover (2)

Fig. 28: compliant box cover (3)

Rarely it is possible to find complex mechanisms based on flexure hinges. One of the most interesting
devices, actually on sale, is the device in Fig. 29 and in Fig. 30. Designed by Mr. Villa, it is a drafting compass
where the kinematics chain is monolithically manufactured in glass-fibre-reinforced polyamide. The
conception of the whole drafting compass is very different from the traditional design: in this case it is clear
that the design process to reach the needed kinematics is based on flexure hinge approach and is not the
attempt to copy a well-knew kinematics replacing traditional joints with flexure hinge. The result is a very
elegant, functional and robust design: the drafting compass has stood until 1 million of cycles without
failure. Another interesting feature is the sliding joint manufactured in a different way: in this case the joint
is not manufactured with flexure hinges but it is a passive joint, using the definition presented by Howell
in [3] (page 183). The Villas drafting compass shows very interesting characteristics and can be useful to
amplify the flexure hinges use.

Fig. 29: Villa's drafting compass

Fig. 31: bistable switch

Fig. 30: Villa's drafting compass (details)

Fig. 32: bistable switch (detail)

10 / 126

Another very impressive device is the bi-stable switch, presented by Howell [3] (Fig. 31). The kinematics
seems very simple but it is not true. In fact the device has two positions of stable equilibrium, shown in Fig.
32, not a standard feature for every mechanism. It is not important to introduce here the explication of the
bistable behaviour (function of the linkage kinematics and the stiffness of every hinge) but this exemplum
shows that it is possible to realize devices with flexure hinges with uncommon behaviour.
In Table 3 there is a resume of the most interesting plastic materials used in flexure based mechanisms:
note that often (always considering the exempla in this paragraph) the flexure hinges are used expressly
over the yield strength. For this reason it is not possible to approach the stress study with traditional
approach. Some interesting advises in studying plastic hinges are provided by Howell [3].
Table 3: plastics for flexure devices




Polyethylene (HDPE)



Polyamide (Nylon 6.6)



Polypropylene (PP)



11 / 126

2 Fundamentals of flexure based mechanisms

Until some years ago there were a lot of papers presenting designs where devices with flexure were used in
different fields. However, there were not publications collecting the basic rules for designing this kind of
mechanisms. In the last years at least three important publications came to the attention to the scientific
community. All three books were published around the year 2000, but the approaches to the compliant
mechanisms are different.

2.1 Howells approach

The first book to appear is the Howell book [3], published in 2001. Howell approaches the study of the
compliant mechanisms paying more attention to the mechanisms with distributed compliance than to the
mechanisms with flexure hinges. His most important contribution in developing a theory for the analysis
and synthesis of compliant mechanisms is the introduction of the pseudo-rigid-body model. The pseudorigid-body model transforms every compliant link in the mechanism (flexure hinges or not) into one or
more one-DOF revolute joints. Every revolute joint is equipped with torsion stiffness, even under the
assumption of large deformations. In this way it is possible to obtain a model to be studied with classical
principles of rigid-body mechanics. To explain the approach, it is possible to observe the three compliant
mechanisms in Fig. 33 a), Fig. 33 b) and Fig. 33 c), which have almost the same pseudo-rigid-body model,
shown in Fig. 33 d). The difference between pseudo-rigid-body models will be the position of the springs.
The device is the Hooken straight-line mechanism (see [3], pages 276-278), where P follows an almost
straight line for majority of its path.




Fig. 33: Hooken mechanism

12 / 126

This exemplum shows that the replacement of a rigid body mechanism with a compliant mechanism (and
vice-versa) can be performed in different ways. This exemplum also clarifies that in the opinion of Howell,
the flexural pivots are only one of many different ways to replace traditional pivots.
Howell tries to find some elementary components of a generalized compliant mechanism: different types of
compliant segments are presented, each with a specific pseudo-rigid-body-model (see Table 4). The flexure
hinges, in this approach, are only one of many different compliant segments. In general the most of the
compliant mechanism members presented by Howell undergo large deflections and the hypothesis of small
displacements is not valid.
Table 4: compliant segments (Howell)

Small-length flexural pivot

Vertical force at the free end of a cantilever beam

Cantilever beam with a force at the free and

13 / 126

Fixed-guided beam

Cantilever beam with an applied moment t the free and

Initially curved cantilever beam

Combined force-moment end loading

14 / 126

Initially curved pinned-pinned segment

One of the most critical factors in this analysis is the fact that very few three-dimensional devices are
studied, reducing the application field. In the opposite way, this one is the only publication that diffusely
approaches the study of mechanisms where the hypothesis of small displacements is removed. In addition,
the approach foresees both the synthesis of a mechanisms and its analysis: for this reason, Howells rules
can be useful for the designer.
Chapter 10 is very interesting because it studies special-purpose mechanisms, with an interesting analysis
on compliant constant-force mechanisms. Special mention goes to chapter 11, where bistable mechanisms
are presented: some fundamentals about stability are introduced and the particular translation in the
compliant mechanism is presented.
Howell has also established a research group, the BYU Compliant mechanisms Research Group [12515],
which is deepening the research about compliant mechanisms with the collaboration of industrial partners.

2.2 Lobontius approach

The Lobontiu book [16] has a rigorous approach: the study is related to almost all the possible flexure
hinges. For Lobontiu a flexure hinge is a flexible, slender region between two rigid parts that must undergo
limited relative rotation in a mechanism due to the presence of at least one flexure hinge ([16] page 1).
Lobontiu divided the flexure hinges into three categories: flexure hinges with single rotation axis, flexure
hinges with two rotation axes and flexure hinges with multiple rotation axes. The attention is focused on
the shape of the flexure hinge: constant rectangular cross section, circular, corner-filleted, elliptical,
parabolic, hyperbolic, inverse-parabolic and secant (symmetric and non-symmetric where it is possible). In
Table 5 there is the shape of almost all the flexure hinges with single rotation axis presented by Lobontiu.

15 / 126

Table 5: different shapes for flexure hinges

a) corner-filleted

b) circular

c) elliptical

d) hyperbolic

e) parabolic

f) inverse parabolic

g) secant

For each hinge the deformations are assumed small within the elastic body (with the small displacement
hypothesis valid) and the compliances are given with an analytic formula. For example for a constant
rectangular cross section hinge (see Fig. 34), the following compliances are evaluated: CxFx, CyFy, CyMz, CzMz,
CzFz, CzMy, CyMy (first index: x, y, z are the displacements, x, y, z are the rotations; second index: Fx, Fy, Fz
are the loads, Mx, My, Mz are the torques).
A comparison between hinges of the same family is performed with respect to the flexibility, the precision
of rotation, the stress limitation and the energy consumption.




Fig. 34: constant rectangular cross section hinge

16 / 126

It is important to focus on the analysis of the hinge DOFs performed by Lobontiu: the single-axis flexure
hinge, studied only in a two-dimensional space, does not have 1 DOF but 3 DOFs; the two-axis hinge and
the multiple-axis hinge are modelled with 5 DOFs and 6 DOFs respectively. In Fig. 33 there is the definition
of the DOFs for a single-axis ([16] on page 24), constant-width flexure hinge, in the opinion of Howell: if the
red link is locked, the mobile rigid link can move in x and y directions and can rotate around z axis. It means
that the white rigid link has 3 DOFs, even if it is necessary to introduce 3 springs to simulate correctly the
hinge behaviour.
flexure hinge
rigid link

rigid link




Fig. 35: DOFs for a hinge in 2D

After that, Lobontiu presents new methods to analyze the statics and the dynamics of flexure-based
compliant mechanisms: each component of a flexure-based compliant mechanism, either flexible or rigid,
will be considered a separate link. Generically, every compliant mechanism is composed only of flexure
hinges since a rigid link is virtually a flexure hinge with zero compliance. The Lobontius efforts are directed
to suggest a way to write an equation system from which it is possible to derive the system behaviour: the
obvious consequence of this analysis is the presentation of a new finite-element formulation for flexure
hinges and flexure-based compliant mechanisms.
The last two chapters of the book are interesting: the second to last approaches some interesting topics like
large deformations, buckling, torsion of non-circular cross-section hinges, composite flexure hinges,
thermal effects, shape optimization, means of actuation and fabrication; the last chapter presents a few
applications of flexure-based compliant mechanisms.
Resuming, the complexity of the approach is very heavy for the designer: the rules presented in Lobontius
book can be easily applied in analysis phase but not in the design phase. The use of FEA becomes
competitive with the Lobontiu approach.

2.3 Heneins approach

The last publication is the Heneins book [17]. The Henein approach is very different from the previous
ones. It is necessary to report some definitions used by Henein ([17], page 6/7):

Elementary flexure articulations: the simplest flexural elements, such as beams, torsion bars,
membranes etc In general it is not possible to obtain with them one mechanical connection with
1 DOF. These articulations do not have, in general, a close equivalent in standard mechanical

17 / 126

Flexure guiding systems: combinations of flexure elementary articulations and rigid links. The
flexure guiding systems are comparable with traditional guiding systems such as revolute joints or
sliders. In general, it is possible to find an equivalent in standard mechanical connections (prismatic
joint, revolute joint, etc).

Articulated structures based on flexure guiding systems (flexible structures): structures composed
of flexure guiding systems and rigid links. In general these structures are comparable with
traditional articulated structures. These structures can have a maximum of 6 DOFs. Some rigid links
are designed to be connected to an actuator, a rigid link is the fixed link and another rigid link is the
output link.

The three definitions represent the ideal way in the design process of a complex mechanism: it starts from
the analysis of the simplest flexure device and then evolutes to more complex devices generally with only
one DOF and finally arrives to structures with 6 DOFs.
In general, Henein ([17], page 19/20) tries to find an equivalence between articulated structures based on
flexure guiding systems and structures based on ideal guiding systems and not an equivalence between
elementary flexure articulations and ideal guiding systems. Simplifying, Henein affirms that it is impossible
to use only a simple elementary flexure articulation to constitute a single DOF connection between rigid
links, consistently with the original analysis preformed on numerous elementary flexure articulations.
In Table 6 there are the elementary flexure articulations introduced by Henein, showing the DOF number
for each articulations. The dotted lines show the locked DOFs, whereas the continuous lines show the
articulation DOFs.
Table 6: elementary flexure articulations

rod: 5 DOFs (square section or circular section)

blade: 3 DOFs

circular hinge: 3 DOFs

torsion bar: 1 DOF (with the right section)

18 / 126

membrane: 3 DOFs with this shape (different

shapes are possible)

bellows: 5 DOFs

Here there is only the evaluation of the DOFs for a simple blade. For the other elementary articulations, it is
possible to perform the same analysis. Henein defines the different stiffness for the bade (see Fig. 36): KM,
KtM, Ktors, Ktrac, KP, KtP, Kcis, Ktcis and then evaluates the stiffness ratios:

KM/KtM>100 if b>10h: the DOF corresponding to the transversal flexion is locked

Ktors/KM1.5: the DOF corresponding to the torsion is free because the two values are almost equal

Ktrac/Kcis>100 if l>10h: the DOF corresponding to the traction-compression is locked

Ktcis/Kcis>100 if b>10h: the DOF corresponding to the transversal flexion is locked

The conclusion is that the simple blade is a connection with 3 DOFs, locking better the others DOFs if b/h
and l/h increase: with respect to the Fig. 36, a connection performed with a simple blade makes 1
translation and 2 rotations possible.

Fig. 36 : stiffness for a simple blade

19 / 126

Note that even the circular hinge has 3 DOFs: the fact that the section of the blade change rapidly is not
enough to have a different number of DOFs with respect to the simple blade. An important consequence is
that changing the shape of the blade (for instance from the simple blade to the circular hinge) does not
help in modifying the kinematics of the device.
The results of this approach are very different from the Howell and Lobontiu approach. Considering a
simple blade in a two-dimensional space:

in the opinion of Howell, if the blade is short, it is a connection with 1 DOF, if the blade is longer,
the evaluation of the DOFs is not so clear;

in the opinion of Lobontiu, the flexible link does not lock any DOF, even if he shows that the
compliances are different;

in the opinion of Henein, in a bi-dimensional space, the simple blade has 2 DOFs, since the other
DOF, corresponding to the torsion of the blade, is an out-of-plane DOF.

Only Henein performs an analysis of the blade in a three-dimensional space and the results have been
previously shown.
Table 7: translation flexible bearings

2-parallel-blades translation bearing

over-constrained-4-parallel-blades linear bearing

4-prismatic-hinges translation bearing

4-circular-hinges translation bearing

20 / 126

Comparing the three approaches, the Lobontiu approach is rigorous (the flexible element has a stiffness in
each direction and it is not infinitely rigid in any direction, for sure) but it does not help to design a
mechanism. In the opposite way, Henein affirms that if there is a ratio bigger than 100 between two
compliances it is possible to consider a DOF locked and the other one free. This approach is very useful, and
it will be used in the following.
After the analysis of the elementary flexure articulations, Henein introduces some flexure guiding systems
with a single DOF (flexure pivots and flexure sliders), describing the main features of all the systems. The
aim is to combine elementary flexure articulations to obtain devices with single DOF, translational or
rotational. At first, Henein presents some flexure prismatic joints (flexure sliders): in Table 7 it is possible to
see these mechanisms. These are the simplest designs. Note that these are not the only single-DOF
prismatic joints but it is possible to imagine different mechanisms (see chapter 3). An exhaustive analysis is
performed on these systems in terms of stiffness, maximum displacement, and motion path of the output
The simplest flexure slider is the 2-parallel-blades translation bearing: if N=0, the stiffness of the devices is

bh 3
(b = blade width / h = blade thickness / l = blade length) and the maximum
displacement is f = adm (adm = maximum stress for the material). Note that the motion of the output

24 EJ y

with J y =

link is a translation but not a rectilinear motion (see Fig. 37). In fact during motion the blades deform in
such a way as to cause a parasitic motion in the direction orthogonal to the needed translation (the result is
a parabolic displacement of the mobile link). Henein shows that it is also possible to evaluate this
displacement: if f is the displacement in x direction and is the displacement in y direction, it is possible to
demonstrate that

3f 2



Fig. 37: 2-parallel-blades translation stage

Furthermore Henein shows that the motion of the output link is a translation only if the actuation force is
applied in a particular point, at a distance from the fixed link equal to half the length of the blades. In the
21 / 126

other way the output link will not only have a translation, but will also rotate. Remember that the stiffness
of the 2-parallel-blades translation bearing depends on the load N on the stage.
It is necessary to highlight a remark about the over-constrained 4-parallel-blades linear bearing (Fig. 38).
The guiding system is frequently used in mechanisms since it looks symmetrical; however it is mandatory to
understand that this design is over-constrained. Comparing the 2-parallel-blades bearing and the overconstrained-4-parallel-blades linear bearing, it should be clear. During motion, the 2-parallel-blades
bearings have a parasitic displacement as previously mentioned. The over-constrained 4-parallel-blades
linear bearing can be imagined such as a composition of two opposite 2-parallel-blades translation
bearings. During motion the parasitic translation of the two 2-parallel-blades translation bearings is
avoided by the symmetry: the consequence is that the blades are obliged to stretch themselves. The overconstrained design has to be avoided because the rigidity of the compliant mechanism, as well the stress,
can change in an unforeseeable way.


Fig. 38: over-constrained-4-parallel-blades linear bearing

About the articulated structures based on flexure guiding system, Henein affirms that it is possible to use
the Grbler formula to determine the DOFs of a planar or a tridimensional structure with n guide with di
DOFs (1in) and b closed loops:

Planar devices:

DOFs = d i 3b
i =1

Tridimensional devices:

DOFs = d i 6b
i =1

Only to clarify the big advantages of this approach, the evaluation of the DOFs of the 4 prismatic hinges
guide ([17] page 146) in a three-dimensional space will be explained. In the device there are 4 articulations
with 3 DOFs and 1 closed loop: with the Grbler formula the DOFs are 6. But a deep analysis shows that the
device has 6 internal DOFs (see Fig. 39), the same number as the result of the Grbler formula. The
experience also shows that the system has 1 DOF: the conclusion is that this device is 1-time overconstrained. In fact the device can function only if the two identical links are in the same plane. In this case
it is not a dangerous problem because it is very simple to respect this condition with EDM machining.

22 / 126

Fig. 39: internal DOFs for a 4-prismatic-hinges translation bearing

It is important to note the following: the term over-constrained condition has been used in two different
ways: during the description of the over-constrained 4-parallel-blades linear bearing and during the
evaluation of the DOFs of the 4-prismatic-hinges translation bearing. If the used term is the same, the
device behaviour is different. In the first case the design is intrinsically over-constrained: the motion of the
output link is strongly affected by the inevitable traction load in the blades. The critical point in this case is
the traction load in the blade, since the blade is obliged to deform in a direction in which it is very stiff. In
the second case the effect of the over-constrained design is lower: in fact, a good EDM machine can
guarantee very easily the alignment of the two arms of the 4-prismatic-hinges guide. During motion, the
blades can deform correctly - they bend - without an undesired load. After all that, the first case has to be
avoided, while the second case is not a big problem: in fact, a lot of the flexure guiding systems are in the
same condition but the functioning does not have any problem.
In addition, Henein introduces some elementary flexure revolute joints (flexible pivots), presenting some
interesting features for each device. Note also in this case that these are not the only single-DOF revolute
joints but it is possible to imagine different mechanisms (see chapter 5). For the most of the pivots, Henein
evaluates the angular stiffness, the maximum angular displacement, the parasitic translation of the mobile
link and the displacement of the instantaneous rotation centre. In the end, Henein compares the different
pivots with respect to:

the displacement of the instantaneous rotation centre: the 2 crossed-separated blades pivot has
the best performances

the stress with the same dimensions: the 2 crossed-separated blades pivot has the lowest stress
with the same displacement (less than 2 times)

to the simplicity of manufacturing: the 2 blades RCC pivot is the simplest to manufacture

For the over-constrained pivot, the same remarks as the over-constrained 4-parallel-blades linear bearing
are valid.

23 / 126

Table 8: flexible pivots

2 crossed-separated blades pivot

2 crossed-not separated blades pivot

2 blades RCC pivot

4 hinges RCC pivot

4 hinges crossed pivot

over-constrained pivot

With the flexure guiding systems shown in Table 8, it is possible to approach the design of more
complicated structures, such as the mechanisms shown by Henein in chapter 10 and chapter 11 [17]. Also,
the second part of the Heneins book is very interesting, providing important knowledge about wire EDM
and the materials to be used in flexure based mechanisms.

24 / 126

In conclusion, Heneins approach is very innovative and different from the other ones. It is recommended
to adopt this approach during the design of metallic flexure based mechanisms, but this approach could
also be used for other materials without validity loss. The power of this approach is the generality and the
simplicity of the models. Henein favours the simplicity of the model instead of the exactitude: in this way it
is possible to apply it in the design phase, providing the instruments to foresee the device behaviour.

In this thesis the Henein approach will be the landmark. The discovery of the Henein approach came late
during my PhD: for this reason some devices were designed without this knowledge. In this case I will try to
review these devices with respect to Heneins guidelines. Some of the others mechanisms were designed in
collaboration with Mr. Henein, some were designed after meeting him. Therefore the design is based on his
point of view.
Also the terminology will be changed, adopting the division in elementary flexure articulations, flexure
guiding systems and articulated structures based on flexure guiding systems: due to this, the term flexure
hinge will be neglected and replaced with elementary flexure articulation.

Fig. 40: butterfly pivot

At this point, it is possible to observe once again the butterfly pivot (Fig. 40), designed by Mr. Henein ([5]).
Since it is composed of four 2-blades RCC pivot in series, there is not the problem to evaluate if the blades
are flexure hinges or not. The device can be named flexure guiding system, with respect to Heneins

25 / 126

3 Traction/compression machine for microdiffractometer

3.1 Introduction
This work was originated by a collaboration between the Mechanical Draw Group and the Chemistry for
Technology Group at the University of Brescia, thanks to a request from prof. Laura Depero and prof. Elza
X-ray diffraction is a well established technique for residual stress measurements: for instance, the strain in
the crystal lattice can be measured, and the residual stress calculated, assuming its linear elastic distortion.
The great advantage of X-ray employment in stress evaluation is the non-destructiveness of the radiation.
The great interest in the XRD study of in-situ mechanical/structural properties is confirmed by several
papers lately published [18]. The development of high intensity synchrotron X-ray sources has significantly
shortened the required exposure time for acquisition of X-ray diffraction patterns, thereby enabling real
time monitoring of structural changes as a function of external stress (strain), as well as it introduced a
significant reduction in the beam spot size [19].
But almost all the past studies refer to Synchrotron applications. Today, the improvement of laboratory
sources makes possible to perform experiments which enable in situ microstructural analysis under applied
stress. For this purpose it is very important to design new instruments that permit to plan, set and develop
The aim of this project is to give evidence of the design and testing phases of a traction/compression
minimachine, suitable to be installed in the commercial microdiffractometer Rigaku D-MAX Rapid (Fig. 41)
located into the Chemistry for Technologies Laboratory at the University of Brescia. This device will be used
to study the behaviour of small specimens under well-known stress conditions, by way of the diffraction

Fig. 41: Rigaku D-MAX Rapid (photo)

The requirements of the device were defined as follows:

it must apply a load of 100 N (traction and compression) with a displacement of 2 mm as regards
to the zero position;

it must avoid interference with the standard operations of the microdiffractometer, especially with
reference to overshadowing the detector area;

26 / 126

it should weight about 1 kg;

it must be fit to allocate a wide specimen variety, ranging from rubbers to metals and from thin
films to wires.

To start the design of the device, a virtual 3D model of the diffractometer was needed. In lack of
information from the manufacturer, the model was developed starting from some digital photos calibrated
by way of traditional measurement tools (Fig. 42).
The main concern was related to the positioning of
the goniometer axis with reference to the
collimator and to the axis of the detector area,
since an error in this positioning would have
dramatic impacts on the possibility to focus the
The detector area, the video camera and the
collimator are the most critical components: during
the mounting/dismounting phase it is necessary to
avoid any contact between the device and the
detector area, the camera and the collimator;
during the focusing phase it must be possible to
move the specimen until precisely reaching the
front view of the collimator; during the
measurement phase the detector must avoid
The traction/compression machine was designed
around a 3-axis positioning stage, used to move the
whole device and consequently to focus the camera
on the specimen.
Fig. 42: Rigaku D-MAX Rapid (draw)

3.2 The Bydlo linear bearing

The traction/compression machine (named Bydlo) is based on the compounding of an untraditional
mechanical linkage the Sarrus linkage (Fig. 43) and an untraditional kind of revolute joint the metallic
flexure hinge.
Weight to be limited and bonds with the D-MAX microdiffractometer induced to select the unusual Sarrus
linkage. The out-and-out Sarrus linkage (an over-constrained linkage with a kinematics point of view), used
in the Bydlo device, is highlight in Fig. 43: it is a perfect 3d linear bearing, that is the link 2 can only
translate along a straight trajectory with regard to the fixed link 1.
In order to reduce dimensions, weight and assembly problems, the traditional revolute joints of the Sarrus
linkage were replaced with the flexure hinges.

27 / 126

The design of the Bydlo frame was the most difficult step of the project. The frame has different functions:
it must perform the tasks of being a linear guide, holding the two grips, supporting the motor/transmission
group and interfacing the whole device to the diffractometer. All such goals were achieved by way of a
proper evolution of the Sarrus linkage, developed in order to allow acceptable displacement by way of
flexure hinges.

Fig. 43: the real Sarrus' linkage

Fig. 44: first version of the Bydlo machine

The preliminary design was coped during my master thesis [20], where I approached for the first time the
Sarrus linkage and the flexure hinges. In Fig. 44 it is possible to observe the first Bydlo device.
At that time the knowledge about flexure hinges was not very developed at the University of Brescia and
the translation from Sarrus mechanism based on traditional joints to the flexure-based Sarrus mechanism
was performed replacing directly the revolute joints with nonsymmetrical-circular flexure hinge. The Sarrus
linkage seemed the good solution for the guiding system because it was possible to have a lot of free space
in the desired position. In fact it was necessary to have the entire diffractometer detector free during
specimen traction/compression (see Fig. 42).
Later on, the process to design a flexure-based compliant mechanism is presented. To face the design of
this kind of mechanism, it is useful to replace the flexible mechanism with a corresponding structure based
on rigid links and ideal joints (the pseudo-rigid-body model suggested by Howell in [3]). For the first version
of the Bydlo device, this translation was performed replacing simply the circular hinges with ideal revolute
joints, assuming that the circular hinge had 1 DOF (note that it is not true, as presented in chapter 2). The
second step is to determine the stiffness / compliance of the hinge used in the mechanism. In my master
thesis [20] it was shown that it is possible to derive by analysis a set of simple expressions suitable to
evaluate stiffness and related stresses/strain relationship for single flexure hinges, such values depending
on the geometry of the flexure hinge. Two hinges were compared: the circular hinge and the nonsymmetrical circular hinge. If for the circular hinge the analytic formula was already well-knew, for the nonsymmetrical circular hinge the analytic formula was developed autonomously.
The stiffness is strictly related to the geometry of the hinge. Under this hypothesis, assuming a beam model
with variable section for the flexural hinge to allow the concentration of its properties on the neutral axis,
an expression that represents the compliance of each hinge can be founded. In general the compliance of a

28 / 126

+ z*

flexure hinge is
EJ ( z )

b s (z )
with J ( z ) =

(see Fig. 45 and Fig. 46). Furthermore,

frequently the thickness t of the hinge is negligible comparing with the radius R of the hinge and the
thickness h of the connected bodies and a simplified expression can approximately represent the
compliance and, avoiding secondary motions, the associated stiffness can be calculated as the reciprocal of
the estimated compliance.

The formula for the compliance of a non-symmetrical circular hinge is the following: c =

9R 2
2 Ebt

with following hypotheses: =




<< 1 and = <<

2.45 .

The flexure hinges are also the zones of the mechanism with highest mechanical stress, therefore an
evaluation of the maximal stress MAX can be reached computing the bending torque M and using the
previous simplified expression for the compliance of the hinge, under proper hypotheses, i.e. linear and
elastic material, De Saint Venants assumptions, homogeneous and isotropic material, and so on:

2 2

Fig. 45: circular flexure hinge

Fig. 46: non-symmetric circular flexure hinge


To be exhaustive, the formulas for the circular hinge are the following: the compliance is c =

9R 2
2 Ebt

with the hypotheses =


, valid


<< 1 and = <<

2.45 .
2 R

The nonsymmetric circular hinges were selected due to better compliance (i.e. a larger displacement for
the Bydlo linear bearing) with respect to widely used symmetric circular hinges.
Evaluating the behaviour of a single kinematics chain in Sarrus linkage, one of the most critical problems
was the high stress level in hinge A (see Fig. 47): in fact the angular displacement in hinge A is the double
with respect to the angular displacement in hinge B. A new kinematics chain was introduced (see Fig. 48) to
obviate this problem: in this case the new linkage has 1 DOF and 8 hinges. Every hinge has the same angular
displacement as the hinge B in Fig. 47. The conceptual hinges arrangement in Sarrus linkage, as shown in

29 / 126

Fig. 44, evolved into the final arrangement as shown in Fig. 49, in order to cope with the basic differences
between flexural and standard hinges. Note once again that at this time it was not clear the DOF evaluation
in a flexure based mechanism. A flexure hinge was considered like an articulation with 1 DOF in the plane;
the out-of-plane behaviour was not so clear but it was decided to consider the hinge such as an articulation
with 1 DOF also in the space.

Fig. 47: scheme of one arm of the Sarrus' linkage (version of the master thesis)

Fig. 48: scheme of one arm of the Sarrus' linkage (definitive version)

Two solutions were possible, as it is possible to see in Fig. 49 and in Fig. 50, but it was impossible to decide
evaluating only the model with rigid links and ideal joints, because the two designs have the same scheme.

Fig. 49: the Bydlo frame (version 1)

Fig. 50: the Bydlo frame (version 2)

30 / 126

The use of FE Analyses helped to choose the final arrangement, which was aimed to reduce stresses and
increase displacements. Because of the availability of a not very performing software it was impossible to
verify a complete model of the device. For this reason it was necessary to analyze a partial device. As
shown in Fig. 53 and Fig. 54, the comparison between the two solutions in Fig. 51 and Fig. 52, was
performed with FEA. Because the solution 1 was slightly stiffer than solution 2, it was decided to use
solution 1.

Fig. 51: partial model of the version 1

Fig. 52: partial model of the version 2

Fig. 53: FEA results for the version 1

Fig. 54: FEA results for the version 2

3.3 Partial models

In order to validate and calibrate the virtual model two partial physical models of the Sarrus linkage were
manufactured. On such physical models some test were performed to check the stress/strain plot, the
yielding limit and the fatigue endurance of the designed flexure hinges.
The study of the behaviour of the two compliant mechanisms with flexure hinges was subdivided in the
same sequential phases such as the study of the whole structure of the machine: the definition of a
pseudo-rigid-body model, the determination of the hinge compliances and the analysis of the pseudo-rigidbody model obtained.

31 / 126


Fig. 55: four-bars linkage for static experiments

Table 9: Geometrical characteristics of the first

quadrilateral mechanism
Geometrical parameter

Assumed value


13.0 mm
30.3 mm

Two four-bars linkage with flexure hinges were

functional compliances. The first trial (Fig. 55) is
conducted by means of a specimen compression
till the yield point, while the descent of the test
bench bar is kept quasi-static to preserve the
equilibrium of the system during the whole
process. A first geometrical examination of this
experimental setup shows a correlation between
the final position of the bar AD, identified by the
angular variable f, and the imposed vertical
displacement s; therefore, fixing the other
geometrical parameters, the angular displacement
of the bar AD can be computed as a difference
between its initial and final positions i and f,
furthermore the particular parallelogram structure
of the articulated mechanism implies the same
relative angular displacement for every hinge. The
same geometrical analysis shows also that the
beam AB is subjected to a transversal
displacement, thus a veil of lubricant was used to
reduce the metal-metal sliding friction between
test bench and specimen.

The evaluation of f is the following: f = arcsin sin i


The flexure hinge consists, as shown in Fig. 55, in a non symmetric circular hinge with the parametric values
listed in Table 10.
Table 10: Geometrical characteristics of the proposed compliant hinge


Assumed value [mm]


Within a stress range that do not produces material yield, the pseudo-rigid-body model obtained can be
examined with an energetic approach: the whole system can be compared with a prismatic joint consisting
in a vertical spring characterized by a stiffness kvert. Observing that the potential energy Ec stored in a single
hinge is a quarter of the energy Emec stored in the equivalent prismatic joint ( Ec = 0.5k 2 ,

E mec = 0.5k vert s 2 ), the expression of kvert can be deducted from the evaluation of f:

32 / 126

4k i arcsin sin i

k vert =
; therefore the statics of the system can be described by the
approximate statics of the prismatic joint: P = k vert s .
The proposed model should be validated and calibrated by means of experimental tests, thus a specimen
was realized in Al 7075 T651 (Ergal), an aluminium alloy with optimal mechanical properties, using wire
erosion technology in two passes: the first is a roughing-out that releases the residual stresses in the
material while the latter is a finishing; finally the specimen was anodized for a protection against corrosion.

Fig. 56: Experimental curve obtained with a quasi static load process

The experimental curve represented in Fig. 56 can be perfectly superimposed, in the linear domain, to the
diagram of expression P = k vert s ; therefore the proposed approximations do not limit the validity of the
model, furthermore the comparison, in the linear domain, with a least square line and with finite element
analyses shows a good correspondence between numerical and experimental models. Finally the
calculation of MAX with the proposed formula appears too conservative, as a matter of fact the forecasted
yield point is really lower than the experimental results; therefore a wider measurement campaign should
be performed on a congruous number of specimens to clear this incongruity.
Besides just described static, or quasi static, tests, also dynamical and fatigue analyses were carried out
with a four bar mechanism analogous to the previous one provided with a vertical mass adequately
dimensioned to be submitted to mechanical vibrations by means of an electro-dynamical shaker rigidly
connected to the link CD of the sample. Initially the specimens are subjected to a frequency sweep to
identify the dynamical characterization of the system, then they are stressed with oligocyclic fatigue loads
according to the forecasted applicative situations. Due to the displacement sin imposed by the shaker to
the link CD, a movement sout of the link AB can be observed; assuming a Fourier expansion for sin, the
resulting output motion sout is not, generally, phased with the input motion, hence the relative motion
between CD and AB constitutes the dynamical load for the four bar mechanism. The scheme of this second
four-bars mechanism (Fig. 57), with the geometrical characteristics listed in Table 11, is different to the
previous one in order to limit transversal accelerations, i.e. secondary motions, of the calibrated mass; as a
matter of fact the morphology of the second mechanism offers transversal motions with a lower order of
magnitude than the previous sample, assuming the same relative displacement between AB and CD.

33 / 126

Table 11: Geometrical characteristics of

the four bar mechanism for dynamical


Assumed value

Fig. 57: Four bar mechanism for dynamical loads

The proposed system can be dynamically associated to a simple prismatic joint realized starting from a
fixed connection and removing it with a compliance, therefore it can be described by a 1 DOF model with a
relative natural angular frequency 0 =

k vert
, or its natural frequency f0. Then the equivalent mass can be

valued as the sum between the calibrated mass and the mass of the link AB, while the determination of the
equivalent stiffness kvert can be achieved with a procedure similar to the one used for quasi static analysis:
known the stiffness of a single compliant hinge (1.230), the dependence of kvert on the displacement s is
neglected for the sake of simplicity by means of a proper choice of the displacement s as a mean value of
1.5 mm in presence of a maximum value of 2.5 mm. Therefore, the behaviour of the mechanism can be
described by the equations


+ (2fc )



4me f

) + (2fc )

2 2

and tg ( ) =

2me cf
k vert
4 2 k vert me f 2 + 4 2 cf 2

which represents, respectively, the modulus and the phase of the transfer function of the system, while the
graphical representation of sout/sin is depicted in Fig. 58.

Fig. 58: modulus of the theoretical transfer function

Fig. 59: modulus of the experimental transfer function

The experimental sweep was carried out within a 01000 frequency range with a maximum acceleration
equal to 0.5 g. Then the transfer function, measured from the accelerometer on the translating plane of the
shaker to the other accelerometer on the test mass on the specimen (Fig. 57), is depicted in Fig. 59 and
compared with the forecasted diagram in Fig. 58, the result is a good correspondence between the
34 / 126

experimental and the theoretical approaches, particularly in the peak at the frequency f0 equal to 41.3 Hz,
therefore the model can be validated. The sample is realized in a metallic material, hence the damping
coefficient assumes a low value, as a matter of fact a calibration of the model should be implemented to
further reduce the value of the coefficient c in the evaluation of sout/sin. Moreover the presence of a low
frequency peak in Fig. 59 can be attributed to the motion controller, while the high frequency peak is
associated with the second natural frequency of the system; however a further study of this second peak
was not performed because the application range for the four-bars compliant mechanism is expected to be
at a lower frequency.
Oligocyclic fatigue tests was executed on the same specimens used for the frequency analysis, exposing
them to a sinusoidal excitation of input/output bars with an amplitude of the movement equal to 2 mm, a
frequency of 30 Hz and over the expected lifetime of 12000 cycles; the theoretical expected results should
be negligible in agreement with the ASM Specialty Handbook [21] for the aluminium 7075 T651.
Table 12: 8 Oligocyclic fatigue test on two specimens

s [mm]



The two samples A and B were subjected, as listed in Table 12, to different cyclic movements with semiamplitudes s at a frequency equal to 30 Hz: the specimen B showed a residual plasticization in the end of
the test with a visible yield of the material confirmed also by an additive analysis of both samples A and B.
Each specimen supported the number of cyclic excitations, but they disappointed the forecasted damage,
probably because of the electro erosion process in presence of thin sections, but further studies should be
performed to confirm this intuition.

3.4 The traction/compression machine

Experience was gained also in order to cope with the manufacturing problems of very thin cross areas. In
the end, the manufacturing of the whole Bydlo frame was performed; the linkage was made with
aluminium alloy (Al 7075 T651), machined by wire EDM. The small thickness of the hinges requested a
special care during the manufacturing: a first machining phase aimed to relieve residual stresses inside the
block material was foreseen, while the final phase allowed one to achieve the required precision.
The functioning of the machine is the following: a DC minimotor (with gearhead) operates a ball screw by
way of a belt linkage (transmission ratio 1:1). The ball screw induces the displacement of the moving link of
the Sarrus linkage. The motion group is standard, the linear linkage is fully analysed in paragraph 3.2. At
present the device is controlled in displacement but the load control is foreseen as a further development.
The overall mass is 1,5 kg instead of 1,0 kg but the diffractometer structure can stand this load without
Stress and strain of the specimen are monitored by way of two sensors. A load cell (range 400 N) is axially
aligned with the specimen and a DVRT displacement sensor (range 4 mm) is located nearby the specimen.

35 / 126

Fig. 60 and Fig. 61 show the device integrated in the microdiffractometer: during the mounting phase the
goniometer of the diffractometer must be set in the 150 position, while in the 0 position, when the
detector area overshadowing is minimised, it is possible to perform transmission measurements (for
instance with plastic materials samples). Around the 0 position (the device can be tuned between 0 and
150) the measurements are to be performed in reflection mode, taking care not to overshadow the
sensible area.
Finally the requirements were fulfilled, with respect to the maximum load (100 N) and the maximum
displacement (2 mm) required. Fatigue tests are also possible but, since the machine is designed above all
for static loads, it is necessary to be careful to the test frequency.
The device was also calibrated to check the real tolerance range of the sensors: 0.4 N for the load cell and
2 m for the displacement sensor were verified. The DVRT resulted slightly less precise than stated by
data sheet, which may reduce the usability for very short metal specimen (until replacing the sensor). To
face up this issue the designers are evaluating an additional displacement sensor, more performing. To test
the copper wire (paragraph 3.5), the grips were used in a non conventional manner to allow one to insert a
rather long wire: the design of proper grips for small metal wires will be coped in the next future.

Fig. 60: the Bydlo machine

Fig. 61: the Bydlo machine into the diffractometer

3.5 Results of the preliminary tests

In cooperation with the researchers of the Chemistry for Technologies Laboratory at the University of
Brescia, preliminary experiments to test the instrument were performed on wires. The samples were
placed in the diffractometer orthogonally to the primary beam.
The aim of the first experiment was to investigate the change in the fibres orientation as a function of the
mechanical load [22]: the test was performed on a polyamide wire, 0.5 mm in radius and 10 mm in length.

36 / 126

The polyamide, a reinforced component, may be prepared by way of several processing techniques,
involving thermal and mechanical treatments, which impact on its crystalline structure and, hence, on its
mechanical properties. Although there is a large number of studies relating the structure or the properties
of polyamide to the conditions of its orientation and annealing, relatively few works have been published
on the relationship between the mechanical properties and structure/microstructure [23].

Fig. 62 : image of the polyamide sample from

diffractometer before the test
Fig. 63: load effect on the polyamide sample

Fig. 62 shows XRD2 image of polyamide sample collected before the load test. The observed crystal
structure is the phase of polyamide, the most thermodynamically stable crystalline form: it is possible to
observe two crystalline peaks at about 20.3 (1) and 23.3 (2). This sample shows high preferred
orientation (fibre orientation).
The intensity distribution around the diffraction rings on the 2D X-ray patterns can be plotted as a function
of azimuthal angle: it is so possible to evidence the preferred orientation. The diffraction intensity profile
was then measured at several load values. Fig. 63 shows that, as the load increases, theres a change in the
intensity distribution along the Debye ring of 2 peak, inversely related to the crystalline orientation (the 1
peak has the same behaviour). This figure shows the change in the preferred orientation induced by
mechanical load and, in particular, the increase of fibres orientation along the load direction.
The second experiment, with a Cu wire (0.3 mm in radius and 30 mm in length) was performed to follow
the sample deformation during mechanical load. Fig. 64 shows XRD2 image collected on the unloaded
sample: all the Debye rings appearing in the 2D image are attributed to Cu phase.
It is well known that, when a mechanical load is applied, it is possible to follow the shift of a Bragg
reflection that is directly associated with a variation of the corresponding interplanar spacing. The lattice
spacing dhkl of a lattice plane (hkl) which is normal to q vector is calculated, using Braggs law from 2 angle
that corresponds to the maximum diffracted intensity. The strain in the direction q is then given by: = (dhkl
- d0)/d0 where d0 is the lattice spacing of a stress-free plane.
In test phase the output of the XRD diffraction analysis was made available in the end of each load cycle.
To evaluate lattice deformations only the portion of the Debye ring in the equatorial plane (orthogonal to
the applied load direction) for the 420 reflection of Cu was considered. In this case, the relative variation
in d spacing allows one to compute the strain of 420 planes orthogonal to the uniaxial applied load. The
change in peak position as the load increases allows one to determine the wire deformation.

37 / 126

Fig. 65 reports the , imposed by the designed device, as a function of the , evaluated with the
diffractometer. From the slope of the fitting line it is possible to extract the elastic modulus, providing that
Poissons ratio is known. In this case the Youngs modulus E420 =100 GPa of the sample is in accord with
the value of the bulk material, as from literature.

Fig. 64: : image of the copper sample from

diffractometer before the test

Fig. 65: tests on copper specimen

3.6 Analysis with Heneins point of view

mobile link
fixed link


Fig. 67: single component of the Bydlo actual frame

Fig. 66: Bydlo actual frame

A kinematics analysis of the linear bearing based on the Sarrus mechanism was performed with Henein
guidelines: in Fig. 66 there is the linear guide of the Bydlo machine. The blue piece is the fixed link and the
grey piece is the mobile link. To simplify the analysis the DOFs only the system in Fig. 67 was considered. As
shown in paragraph 2.3, the circular hinge is an articulation with 3 DOFs, such as the simple blade. In this
case applying the Grbler formula, with 4 hinges and 1 loop, the result is 6 DOFs. Deepening the analysis
there are 2 internal DOFs and the mobile link has 5 DOFs, because only its rotation around y axis is locked
twice (and the device is over-constrained). In Bydlo kinematics, every orange chain is composed of two
elements such as the four-bars linkage in Fig. 67: the DOFs in the end of the chain (one arm of the Bydlo

38 / 126

linear bearing) are the same as the single element (Fig. 68). At this moment, composing the 2 chains it is
possible to analyze the DOFs of the mobile link: the Grbler formula (two joints with 5 DOFs + 1 closed
loop) shows that the output link has 4 DOFs. The composition is possible with simplicity because the two
chains are orthogonal: the result is shown in Fig. 69. It is possible to conclude that the Bydlo linear stage is
not a linear bearing with respect to the Henein guidelines. This comment can be useful to increase the
device performances.

Fig. 68: output link of the Bydlo linear bearing with the
DOFs of the two arms (actual version)

Fig. 69: DOFs of the output link (actual version)

Later on, two solutions to obtain a real linear guide for the Bydlo machine with simple modifications will be

Fig. 70: solution A for the Bydlo frame

In the first solution, each kinematics chain composing the arms

of the Sarrus mechanism can be replaced with a series of two 4prismatic-hinges guides, as shown in Fig. 70. Each kinematics chain
has 2 DOFs (Fig. 71) and the coupling of the two provides a device
with only one external DOF (Fig. 72). For sure there are a lot of
internal DOFs (each 4-prismatic-hinges bearing has 6 internal
DOFs) and some over-constraint degrees too (for instance both the
kinematics chains lock all the rotations). It should be possible to
evaluate the number of the internal DOFs and of the overconstraint degrees but it is not useful in this case. However it is
important to remind that the solution is over-constrained because
of the problems that this condition can cause. In this application
the internal DOFs are less important because the Bydlo machine
does not work in high frequencies domain.

The second possibility is to use the Sarrus/Henein mechanism, a single-DOF linear bearing. Henein ([17]
page 154) presents his personal vision of Sarrus linear guide. In fact, he uses the three-dimensional
coupling of two kinematics chains to have a perfect linear displacement. The idea at the basis of this design
starts from a traditional 2-parallel-blades guide: to compensate the parasitic transversal motion Henein
introduces another 2-parallel-blades guide in an orthogonal direction (Fig. 73). At this moment the mobile
link has 2 DOFs: Henein adds a second kinematics chain with the purpose to lock the transversal motion of

39 / 126

the mobile link: linking the second chain directly to the mobile link, the mobile link will have a perfect linear

Fig. 71: output link of the Bydlo linear bearing with the
DOFs of the two arms (I solution)

Fig. 72: DOFs of the output link (I solution)

An analysis of the DOFs can be performed with the Grbler formula: with 7 articulations with 3 DOFs and 4
closed loops (see Fig. 73), the structure is 3 times over-constrained. However, since the experience shows
that this mechanism has 1 DOF, it is possible to affirm that the internal structure is 4 times overconstrained. In fact, in this device there are 3 complete 2-parallel-blades translation bearings: each one is
time 1 over-constrained (such as the 4-prismatic-hinges guide in paragraph 2.3), for a total of 3 overconstrained conditions. To find the last one, every 2-parallel- guide blades is replaced by an equivalent ideal
joint, as shown in Fig. 74. The loops 1 and 2 (stage 1 and 2) have the effect to lock all the DOFs except two
translations; the remaining flexures (one 2-parallel-blades guide and a single blade) locks one DOF more,
the rotation around x axis (Fig. 75). The conclusion is that this device is 4 times over-constrained. The big
advantages are that this mechanism does not have internal DOFs (avoiding all the problems due to the
internal DOFs) and that the motion is perfectly rectilinear.

Fig. 73: Sarrus/Henein mechanism

Fig. 74: simplification of the
Sarrus/Henein mechanism

Fig. 75: DOFs of the

Sarrus/Henein mechanism

Adding a blade in the loop 3 (see Fig. 76), the device becomes more over-constrained, but with the
advantage to have two equal kinematics chains: it is possible to manufacture two planar equal piece to be
assembled later. Adding the new blade has the effect to increase the over-constrained condition of the
mechanism: in this case the device is 7 times over-constrained (because all the 3 DOFs locked by the new
blade were already locked).
40 / 126

It is possible to implement this solution also in the Bydlo machine. In Fig. 77 there is a possible design. Note
that the Sarrus/Henein device in Fig. 77 seems different from this one in Fig. 76: in fact in Fig. 77 2-parallelblades guides were replaced by two 4-prismatic-hinges bearing. With respect to the functionality of the
device, there is no difference. The difference will be in the stiffness in displacement direction and in
stiffness ratios of the guide, that is the ratios between stiffness in displacement direction and the stiffness
in the other direction, for instance the stiffness in blade direction (see [17], page 95). In the future, to
choose the new solution it will not be sufficient to verify the kinematics but it will be mandatory to verify
the stiffness of the new solutions, comparing these one with the actual device.

Fig. 76: the Sarrus/Henein mechanism: a little


Fig. 77: the Sarrus/Henein mechanism in the Bydlo


3.7 Conclusions
This work presents a new laboratory device designed to study the change in the structural/microstructural
properties of small components during mechanical loading.
The first tests were performed on wire samples of known materials to verify the instrument reliability. The
results are very encouraging and the researchers are planning future experiments with more complicated
samples (thin film on a substrate, multilayers, etc.).
A collaboration with research group designing a similar device for Synchrotron applications has been
planned to compare the instrumentations and the results for thin film samples.

41 / 126

4 Compliant universal joint

4.1 The design of a compliant universal joint
This work was initially faced in collaboration with Mr. D. Gazzoli, a student of the Faculty of Mechanical
Engineering at the University of Brescia. The main results are collected in his bachelor thesis [24].
The starting point was to translate in a flexible device the kinematics of the universal joint. It was planned
to use it in a strong load case, as a decoupling element between a hydraulic actuator and a translation unit
with hydraulic chariot, as it is possible to see in Fig. 78. In this way it should be possible to avoid an overconstrained solution. The work development has shown the impossibility to use the flexible universal joint
designed due to the exaggerated involved loads: about 700 kN in axial direction. But the study has provided
some interesting results.

Fig. 78: the foresee employment for the compliant universal joint

Considering the traditional universal joint shape, some different solutions were proposed with different
blades dispositions. In fact, the idea was to replace the traditional hinges in the joint directly with flexure
hinges. Considering the two rotation axes in a universal joint, it is possible to dispose the hinges in two
ways: with the blades in the axial direction, such as in Fig. 79 A, or in transversal direction, such as in Fig. 79

Fig. 79: alternative dispositions of flexure hinge to obtain a universal joint

42 / 126

It was immediately clear that the disposition A could be useful to pass on the axial loads but not a torque.
In fact, the blades have a high stiffness in axial direction but a very low stiffness in transversal direction. The
disposition B presents the opposite condition, good for torque but not for axial load. A monolithic design
was presented for both the solutions. Working on the solution A, a more elegant design was derived: it is
possible to design this solution starting from a pipe, with very simple EDM machining. The solution A
obtained from a pipe was named pipe universal joint and the solution B was named wall universal joint.

Fig. 80: compl. un. joint for axial load (sol. A)

Fig. 82: pipe un. joint

Fig. 81: wall un. joint for

torque load (sol. B)

The will to find a flexible universal joint with better properties with respect to the loads led to the
combination of the solution A and B in a new design, the version C. In solutions C, the pipe universal joint
and the wall universal joint are coupled, so as to have the advantages and to cancel the disadvantages of
both solutions. In this way the core, very stiff with respect to the traction/compression load, will support
the entire axial load, while the external part, very stiff to the tangential loads, will support the torque. The
only disadvantage is that the number of the blades is doubled and if all the blades are identical, the
rotation stiffness is double with respect to solutions A and B. One of the most important requirements in
this solution is the alignment between external blade and internal blade: if the 2 blades of the same axis
have a circular profile, it should be very difficult to align the external and the internal joint. With a linear
profile this problem has a low importance. The solution C was named complete universal joint.

Fig. 83: solution C (complete un. joint): combination of solutions A and B

4.2 3UPU parallel kinematics robot

It was decided to apply the universal joint to a parallel robot for vacuum applications. For this reasons it
was decided to use an actuator specific for vacuum application: the piezo actuators were the ideal solution.
43 / 126

To obviate the small displacement of piezo actuators, the a piezo inch-worm was chosen: the actuator, the
PI N-214, has a big displacement ( 20 mm) with respect to the traditional piezo stack. The nanometric
resolution is another important feature. The technical data sheet shows that this kind of actuator is advised
for vacuum applications, in clean rooms and where the magnetic field must be avoided. The PI can provide
the actuator with an integrated encoder and controller for closed-loop operations.

Fig. 84: PI N-215 NEXLINE high-load actuator

The material chosen for the universal joints was Titanium Ti6Al4V STA. The flexure hinges were
dimensioned for a maximum displacement of 4,5, with an axial load of 300N. A fast analysis was
performed on the working space for the robot: the results are shown in Fig. 85, Fig. 86, Fig. 87 and resumed
in Table 13.

Table 13: main features for the parallel robot

displacement in x direction

17 mm

displacement in y direction

17 mm

displacement in z direction

11.5 mm

actuator displacement

20 mm

max loads (Fz, Fz, Fz) on the output


500 N

max torque Mz on the output link

90 Nm

Fig. 85: the 3 linear DOFs robot (A)

Fig. 86: the 3 linear DOFs robot (B)

Fig. 87: the 3 linear DOFs robot (C)

44 / 126

4.3 Dimensioning of the compliant joint for the 3 DOFs robot

Due to its high yield strength, the material chosen for the study is a Titanium alloy for thermal treatment:
the Ti6Al4V STA (+ class). This alloy can be easy found on the market. The high mechanical performances
and the high index of deformability in elastic range E/sn support the choice (see Table 14).
Table 14: mechanical features of Ti6Al4V STA

E [MPa]

r [MPa]

yield strength
sn [MPa]






infinite life
(alternate flexion
load) [MPa]

cycles for the

infinite life
Nfa [cycles]

Then it is possible to start the dimensioning of the two universal joints: in fact the complete universal joint
is composed of two joints, the joint for the axial load and the joint for the torque. For both the joints it was
decided to use the circular hinges but the dimensioning was performed independently for each joint.
Since the pipe joint is turned into the axial load, whereas it is not affected by the torque on the complete
joint, it is possible to conclude that the loads acting on this pivot are: an axial load, equal to the holding
torque of the actuators (Ftot=300 N F=150 N on each hinge of the pipe joint) and an angular deformation
(max = 4.5). The maximum stress is evaluated with the formula max = ax + flex =

F 6 max
A c f bt 2

where c f =

9R 2
2 Ebt


With proper considerations, for instance choosing the percentage of the axial stress with respect to the
total maximum stress in the blade, the geometry of the blades of the pipe joint is completely defined:

t = 0.3 mm
b = 13 mm
R = 25 mm
l = 9 mm

Fig. 88: hinge for pipe joint (3 DOFs robot)

Some FE analysis with the software CosmosWorks were performed to check the simplified dimensioning of
the blades. To begin with, a comparison was performed by analyzing the node values and the element
values provided by the software in order to choose the right dimension of the mesh (Fig. 89).
It was decide to analyze the joint with the intermediate mesh, rounding the node stresses up to the nearest
unit. As it is possible to see in Fig. 90, it was impossible to perform an analysis on the whole pivot due to
the limits of the software Cosmos Works.

45 / 126

Table 15: comparison between different meshes

elements in the

Node stress

Fig. 89: comparison between different meshes

Element stress


Fig. 90: stresses (von Mises) for pipe un. joint

As it is possible to observe in Fig. 90, the stressed area in the pivot is concentrated around the minimum
thickness, whereas the other part of the structure is comparable to a rigid link. It is also interesting to note
that the motion of the rotation axis of the hinge is limited. The results of the FE analyses are very close to
the stresses evaluated with a theoretical approach. In Table 16, there is a comparison between FEA and
Table 16: comparison between theory and FEA (pipe un. joint)


max [MPa]

flex [MPa]

axial [MPa]

To guarantee homogeneity in the elastic behaviour, the same geometry of the pipe joint was adopted into
the wall universal joint, evaluating only a different width b of the blades. The maximum foreseen torque on
the joint is 30 Nm. To perform a simple evaluation of the stress on the wall pivot, the torque was
decomposed in different forces, located in the middle of the hinges. The value of the distance r from the
hypothetic cross of the two rotation axes was chosen in order to limit the overall dimensions of the pivot:
r=40 mm. Then it is possible to evaluate the load T=torque/r=375 N. In the end, since the load condition for
a blade of the wall universal joint is the same as the pipe universal joint, the same formulas were used to
define the dimension b. Imposing a maximum stress due to the force T, the dimension b assumes the value
of 28 mm.
Also in this case, an FE analysis was performed to check the obtained result. Because the geometry of the
blades and the load conditions of the wall universal joint are comparable to those of the pipe joint, it was
decided to use a mesh with 2 elements in the minimum thickness. Also in this situation it was not possible
to perform the analysis of the whole structure because of the unavailability of performing FE software: the
von Mises stresses for analyzed model is shown in Fig. 92.

46 / 126

Table 17: comparison between theory and FEA (wall un. joint)

max, tot [MPa]
flex [MPa]
torsion [MPa]

The comparison between FEA and theory presents good results, also in this case. The main difference is the
transformation of the torque acting on the pivot in a force acting axially on the blades.

Fig. 92: stresses (von Mises) for wall un. joint

Fig. 91: from the torque to the axial loads on the

In the end it is possible to collect in Table 18 the information from the different analyses to obtain the
features of the complete universal joint, used in the PK robot for high vacuum applications.
Table 18: features of the complete universal joint

pipe un. joint

wall un. joint
complete un.

Kf [Nmrad ]

Ka [Nm-1]

Kt [Nmrad-1]

Mf4.5 [Nm]

lmax [mm]

max [deg.]

Kf is the flexional stiffness of the whole pivot for each rotation axis, Ka is the axial stiffness of the whole pivot, Kt is the
torsional stiffness of the whole pivot, Mf4.5 is the needed torque to have the rotation of 4,5 around one axis, lmax is
the deformation in axial direction due to the maximum force (300 N), max is the deformation due to the maximum
torque (30 Nm).

4.4 Analysis with Heneins point of view.

Henein has shown that at first it is important to analyse the DOFs of the flexure hinges mechanism. In order
to have a flexible universal joint it is mandatory to have a flexure kinematics with the same number of
DOFs. In the pipe universal joint the cross link has only 3 DOFs with respect to the fixed link since every
blade has 3 DOFs (Fig. 93 A). The same happens if the cross element is locked and the DOFs of the second
element are evaluated with respect to the cross element (Fig. 93 B). With respect to Fig. 93 A, B, C and D,
the dotted arrows indicate the DOFs locked by the flexures. The result, shown in Fig. 93 C, is that the
analyzed structure has 5 DOFs: only the DOF in the direction of the axis of the pipe is locked. The main
consequence is that this mechanism is not a universal joint. It would be interesting to observe that the 2
monolithic devices in Fig. 93 C and Fig. 93 D have the same DOF number.
47 / 126

fixed link

cross link

output link

Fig. 93: DOFs for the pipe un. joints and comparison with the simple rod

Fig. 94: DOFs for the wall universal joint

Fig. 95: DOFs for the complete universal joint

Performing the same analysis on wall universal joint, the result is different: the mechanism has 3 DOFs
since the rotation around the longitudinal axis and the 2 translations in the plane of the blades are locked

48 / 126

(see Fig. 94). But the analysis on the complete universal joint provides a very different result: the pipe
universal joint and the wall universal joint in a parallel disposition provide only 2 DOFs on the output link,
as it is possible to obtain adding the locked DOFs shown in Fig. 93 C and in Fig. 94. Only the complete
universal joint has the same kinematics as the traditional universal joint.
Focusing the attention on the disposition of the flexible blades, the blades 1 and 2 in Fig. 96 A are linked to
fixed link of the complete universal joint and each one is linked to one of the two cross elements, in the
intermediate position in the joint.
blade 1
blade 2

blade 1


blade 2

output link

cross link 1

cross link 2
Fig. 96: a different possibility for the universal joint

The positions of the blades remind one of the crossed-separated-blades revolute joint introduced by
Henein. Using four of these joints it is possible to realize a universal joint, with 2 DOFs. It is possible to
affirm that the complete universal joint approaches the solution with four crossed-separated-blades
revolute joints but only the last one translates the traditional universal joint in a flexible device.

Fig. 97: traditional universal joint

Fig. 98: un. joint with a single rev. joint for each axis

It is necessary to add that the last solution presents the same characteristics as the traditional universal
joint with respect to the over-constrained design. In fact, two revolute joints would not be necessary to the
kinematics. With respect to Fig. 97, it would be sufficient to have only one revolute joint between link 1 and
link 2, and one between link 2 and link 3, as shown in Fig. 98.
49 / 126

Fig. 99: Henein's universal joint

Henein proposes in [17] the solution shown in Fig.

99. The approach is completely different: the
starting point is the big central blade. The solution
is an original universal joint which looks very
different from the traditional one. It is important to
observe that in this case the priority is in the
outgoing kinematics and not on the blind
translation of a well-known device. Considering only
the central blade as connection between the two
blocks, the mobile block has 3 DOFs. Considering
only the link orthogonal to the central blade
connecting the two blocks, the mobile block has 5
DOFs. The advantage of the real design is that the
central kinematics chain locks exactly the desired
DOFs of the big blade, freeing the other ones. It
would be interesting to compare the different
solutions with respect to the stiffness, the
dimensions and the maximum displacement

4.5 Use of the flexible universal joint

Fig. 100: balance at IMGC in Turin

Fig. 101: balance at IMGC in Turin (zoom)

During the research in the literature, it was discovered that the flexible joint had already been used in some
application, often in a version with incorrect DOF number. In Fig. 100 and Fig. 101 there is the balance for
high precision measurements designed by the IMGC (Istituto Metrologico Gustavo Colonnetti - C.N.R. Torino), for calibration of masses between 10 kg and 50 kg, with an accuracy of 10 mg. In this application
the joint has the task to allow the alignment of the weight (such as the traditional knives): in this way the 5
DOFs do not disturb the functioning since the joint works only in axial direction. Note that the blades were
manufactured with silicon steel 52SiCrNi5, subjected to hardening at 850 C, quenching in oil and
tempering at 50C.

50 / 126

5 Mechanism for space application: the LISA PAAM

As mentioned in the introduction, during my stage at the CSEM I was involved in some projects: one was
related to the design of the PAAM device. In fact, the CSEM is actually concerned to design the PAAM
(Point Ahead Angle Mechanism) for the LISA Project by the European Space Agency. The LISA Project
foresees three identical spacecrafts, located at 5106 km apart forming an equilateral triangle (see Fig.
102). LISA is a giant Michelson interferometer placed in space. Due to the enormous distance between
spacecrafts, the incoming and the outgoing beams must have different direction in order for the two
spacecrafts to be optically locked. The PAAM has the task to compensate the misalignment: as shown in
Fig. 103, the device has to orient a mirror around an axis (the z axis) with a maximum angular displacement
of 412 rad.
The specifications for this device are very strong. At first, there are very strong requirements in stability of
the rotation axis in the xy plane (see Fig. 103): the LISA project is directed to the detection of the
gravitational waves, by measuring the optical path between two proof masses in two spacecrafts. The
measurement bandwidth shall be from 10^-4 to 1 Hz. The required longitudinal pathlength stability is

1,5 pm
in the same frequency range. The stability requirement was translated
defined as
1 +

in a static requirement of a maximum parasitic displacement of the axis at 412 rad of 21.786 pm (= 21.8
ESA have required also the redundancy in the actuation: if one actuator fails, the other one has to
guarantee the mechanism functions.
Moreover, the device has to support the lunch without locking devices, with a first eigenfrequency higher
than 150 Hz. Last but not least, the allowable volume is very small - 40 x 60 x 60 mm3 - and the maximum
weight has to be 100 g.

Fig. 103: the mirror for PAAM LISA

Fig. 102: the LISA project

5.1 Choice of the actuator and the sensor

To respect the strong requirements about the stability it was decided to adopt a step strategy. It was
demonstrated that with a motion of the actuators with steps of some nanometres the requirements in
term of stability were satisfied. Because of the small steps, the piezoelectric technology seemed the most
promising. In particular, it was decided to not use the piezo stacks that require a specific electronic to

51 / 126

control the steps, with thermal problems due to the presence of a continuous alimentation. It was decided
to use a piezo inchworm, which is designed to make small steps with a high precision. The working principle
is shown in Fig. 104: in every unit, the piezo A has the function to give the piezo B in contact with the
mobile bar; then the piezo B (shear piezo, for which the deformation is not in the axial direction) deforms
pushing the mobile bar. Because there are two sets of piezos, while the set 1 moves the bar, the set 2 is not
in contact with the bar. When the set 1 has finished the step, the set 2 acts to hold the bar. After that the
set 1 has released the bar, it is the turn of the set 2 to deform and to push forward the bar. In this way it is
possible to have a very big displacement, bigger than the shear deformation of a single B piezo.


Set 1

Set 2

Fig. 104: the piezo inchworm

Table 19: actuators for LISA PAAM




PI N-310.01

Piezo bimorph drive

NEXACT PiezoWalk

35 mm

40 mm

Operating range (stroke)

~ 40 um

Absolute accuracy

~ 50 nm



10 nm

5 nm

Go-to-position speed

~ 500 nm/s

12.5 mm/s

10 mm/s

Maximum Power dissip.

< 10 mW



Volume (mm3)

30 x 30 x 30

22 x 11 x 20

25 x 25 x 12



20 g

50 g

No launch locking


7.3 N

10 N



Non magnetic or no
ferromagnetic materials
Vacuum capability

10-7 torr



Operating Temperature

+10 - +40C

-20C - +70C

0 to 50 C

Two suppliers were chosen, Piezolegs and PI. In Table 19 there are the most important features of both the
actuator: the only difference between the two is that the Piezolegs has one set of walking legs and the PI

52 / 126

has two opposite sets of walking legs (such as the Fig. 104). The Piezolegs Piezo Bimorph Drive was chosen
because of the smaller dimensions and the availability in a reasonable time. The PI N-310.01 was
maintained as second choice.
About the sensor, it was necessary to find the most performing sensor on the market, with the smallest
dimensions as possible. With the required resolution it was mandatory to look for a capacitive sensor. In
the end of the trade-off the Micro-Epsilon sensor was chosen: the most problem is that this sensor is not
available for space application. In this moment the CSEM is verifying the possibility to have a custom
version. It seems possible because the capacitive sensor exploits the geometry to make a measurement,
independently from the material. The second selection is the Queensgate sensor: the most disadvantage of
this solution is the volume, very high with respect to the specs.
Table 20: sensors for LISA PAAM

Sensor range
Sensor Linearity
Sensor Temp. stability (zero drift)
Sensor Temp sensitivity (scale)
Electronics Temp stability
Sensor dimensions
Target diameter
Space heritage

50 m
0.1 m
0.5 nm
0.06 m/C
11 ppm/C
< 0.01%FSO/C
< 0.005 m/C
6mm x 12mm

SPNS 1100
240 m
<0.2 m
18 nm 0.0077%FSO*
---9.5mm x 22mm

* specified as noise of sensor

Then it was decided to couple two identical sensors, in such a way as to have a differential measurement,
to provide a better accuracy and at the same way.

5.2 Preliminary asymmetric design

Fig. 105: pivot for first version of LISA PAAM

Because of the very strong requirements in the

quality of the rotation around the ideal axis, it was
decided to suggest a flexible pivot with theoretical
fixed rotation axis. As shown in Henein [17], almost
every flexible pivot has a parasitic displacement of
the rotation axis during motion. Nevertheless it is
possible to find some solution in which the pivot
axis is fixed during motion. Henein has presented
the solution shown in Fig. 105: the pivot is isostatic,
with respect to Heneins point of view.

In fact during motion every blade deforms with an S shape and is not stressed by a traction/compression
load. The consequence is a parasitic vertical displacement of the mobile link (Fig. 106). Because there were
no strong requirements about the displacement in this direction, this pivot was judged acceptable.
53 / 126

Fig. 106: parasitic displacement of the pivot

The first version of the complete device is shown in Fig. 107 and Fig. 108: the flexible pivot is moved by a
tangential force applied through a long flexible blade. The length of this blade was chosen to avoid a too big
parasitic load in the direction orthogonal to the mirror: in fact during motion the blade deforms, applying a
reaction load in this direction. Because the main pivot is not infinitely stiff, the blade reaction has an effect
on the axis rotation of the pivot, displacing it. This effect has to be minimized (21.786 pm at 412 rad).


Fig. 108: LISA PAAM (I version) - B
Fig. 107: LISA PAAM (I version) - A

18 mm
actuator (B)

actuator (A)


38 mm

Fig. 109: LISA PAAM - the actuation lever

54 / 126

To have the redundancy in the actuation, a lever was added: in normal functioning only the actuator B is on
and the redundancy lever pivots around the flexure hinge connecting itself and the other actuator. If there
is a failure of the nominal actuator, the redundant actuator has to have the necessary stroke to maintain
the device functionality. With this design the maximum required displacement for the actuators is 66 m.

5.3 Symmetric design

Fig. 111: LISA PAAM - last version - B

Fig. 110: LISA PAAM - last version - A

ESA requested to double the pivot because the single pivot seemed too weak with respect to the launch
conditions and the configuration was judged not optimal with respect to the thermal requirements. The
ESA was contended (Fig. 99and Fig. 100) but doubling the blades causes a different mechanism behaviour.
In fact the new pivot becomes over-constrained: with respect to the Fig. 112, during motion the mobile link
should go up because of the flexion of the top blades but it is obstructed by the blades of the bottom pivot.
The result is that during motion the blades are stressed by a traction load, giving rise to a non linear
behaviour of the pivot. In this case, the over-constrained solution was accepted by the CSEM because the
displacement is very small and the behaviour can considered linear.

Fig. 112: over-constrained pivot for PAAM

55 / 126

Fig. 113: the monolithic pivot for LISA PAAM

Because of the stability requirements, it was

decided to realize the mechanism monolithically:
the result is shown in Fig. 113, where the double
pivot and all the flexible blades are manufactured in
the same block. In Fig. 115 it is possible to see the
redundant lever. The first version of the redundant
lever (see Fig. 114) had some design problems: in
fact the old version of the redundant lever did not
have the right DOFs, causing some doubts about
the functioning and the resistance during launch.
The stability problems are solved replacing the rods
between actuators and redundant lever with
blades, using on the lever a pivot able to support
also radial loads and gluing the actuator movable
links to the redundant lever.

Fig. 115: redundant lever (last version)

Fig. 114: redundant lever (old version)

5.4 PAAM dimensioning and material choice

This one is the occasion to present a different way to design a flexible device. This way does not foresee the
introduction of the pseudo-rigid-body model but considers the bonds and the load conditions of each
blade, as it will be clear later on.
At first, the most important step is the design of a device able to cope with the requirements in terms of
DOFs. This phase cannot be very easy but Heneins approach, shown in paragraph 2.3, is very useful. The
goal of this phase the synthesis phase - is generally the arrangement of a variable number of elementary
flexure articulations in such a way as to obtain a structure with the right mobility. Note that it is not very
important, in this step, the shape of the blade: it is possible to use for the first version only simple blades
56 / 126

with rectangular cross section constant along the blade. It is the direct consequence of the consideration
that the circular hinge has the same number of DOFs than the blade with constant rectangular cross
section. With respect to the ESA pivot, it was decided to use the pivot already introduced, since it does not
have a parasitic displacement of the rotation axis during motion.
The second step is the choice of the dimensions of the blades in the flexure structure. A preliminary
analysis can be performed considering the load condition of each blade. Henein in [17] collects almost all
the formulas necessary for the first dimensioning of a blade: different load cases are considered and for
each one the stiffness of the blade and the maximum displacement are evaluated. In this case, under ideal
conditions, the pivot is loaded by a torque Mz and two forces Fx and Fy (see Fig. 116). The loads were
considered separately: for instance, the torque was divided on the three blades, as it is shown in Fig. 117.
At this point it was possible to consider each blade separately.



Fig. 116: considered loads on the pivot

Fig. 117: torque Mz divided on the blades

The load condition shown in Fig. 119 was considered: the deformed blade shape is an S shape, that is the
free end does not rotate under load. For the over-constrained pivot, the right load condition should be that
one in Fig. 120, because the blades are loaded axially, but for small displacements the traction load can be
neglected and the 2 cases are equivalent. The stiffness of the blade in this condition is K cis =

Jy =

12 EJ y


bh 3
and the maximum displacement is f cis = adm , where adm is the fatigue strength for the

material (in Fig. 118 there is the legend for stiffness analytic formulas).



Fig. 118: blade


Fig. 119: load case for a blade
of the pivot

Fig. 120: real load case for

a blade of the pivot

57 / 126

To evaluate the radial stiffness of the pivot, only one active blade per pivot was considered. The blade can
be considered with the same load condition as the previous case but with a different stiffness because the
load is in a different direction with respect to the previous case: K cis

corresponding maximum displacement f cis =

adm l 2

12 EJ x
and the


Fig. 122: load case for a blade

of the pivot (radial direction)

Fig. 121: radial load on the pivot

To dimension the remaining flexures, the driving rod and the two pivots of the redundant lever, the scheme
shown in Fig. 112 was adopted.
main pivot


Fig. 123: scheme for flexures of the redundant lever

Fig. 124: load case for the

driving rod

In Table 21 there are all the formulas used for the dimensioning of the flexures: for each case in Table 21, h
is the thickness of the blades and b is the width. As previously mentioned, the radial load on the flexible
main pivot is due to the deformation of all the flexures placed in the device, in particular the flexure A, the
driving rod. Therefore, studying the flexure A with the scheme in Fig. 124, it is possible to derive the load F
acting on the main pivot. A fast analysis shows that, once M1 is evaluated (see Table 21), M 2 =

F1 = F2 =


3M 1

58 / 126

The dimensions of all the flexures were chosen trying to have the lowest angular stiffness of the pivot with
the maximum radial stiffness.
Table 21: schemes for flexures in redundant lever

k =

Jy =

4 EJ y

bh 3
; adm = adm
2 Eh


bh 3
2 l
adm = adm

k =

EJ y

; Jy =

k =

8 EJ y

; Jy =


adm =

adm l

bh 3

2 Eh

All the formulas were introduced in a Matlab file to derive the ideal dimensions of flexure blades in the
device. Considering the copper-beryllium, the results for the pivot blades are the following:

b=10 mm

h=130 m

l=2.3 mm

Note that these dimensions do not change if the material changes. In fact, the goal is to limit the radial
displacement of the main pivot but the disturbance in the radial direction is caused by an elastic load, the
deformation of the driving rod that gives rise to the rotation of the main pivot. If the material changes, the
stiffness of the main pivot (due to the Youngs modulus) increases, for exemplum, and the stiffness of the
driving rod increases as well, increasing the radial load (because the angular displacement of the lever is
constant). Concluding, with a stiffer material the disturbance increases and the radial stiffness of the main
pivot increases as well: the displacement of the pivot in radial direction does not change.
Table 22: comparison between analytic evaluation and FEA (main pivot))


FEA (simple blade model)

Matlab (with
shear effect)
ka = 192 Nm/rad

ka =217.11 Nm/rad

(Matlab with

kr = 2.86E+09 N/m

kr = 4.77E+07 N/m

kr = 7.88E+07 N/m


59 / 126

Then the following step was the check of the stiffness with a FEW software. In fact the introduced
hypotheses helped during the dimensioning but it was necessary to verify it. It was observed during the first
analyses that the FEM stiffness in radial direction was lower than that one evaluated with the analytical
formulas. As it is possible to observe, the blade is very short with respect to the width: for this reason the
shear effect was added in the analytic evaluation of the radial stiffness of the pivot. The results in this case
were similar and it was decide to keep the dimensions of the blades (Table 22).
An ideal FEM model has been constructed with identical meshing in all six blades in order to avoid inherent
errors that exist from differences due to meshing. Automatic meshing has been avoided since this would
cause parasitic effects due to the difference in apparent stiffness in each blade. Some figures of the FEM
analyses are reported here.


Fig. 125: model of the main pivot for FEA

Fig. 126: effect of Mz on the pivot

With the previous 3D model the stiffness matrix of ideal flex pivot was evaluated. In Table 23 there is an
elaboration of the stiffness matrix, where the displacements of the centre of the pivot are presented under
different loads.
Table 23: 3D displacements of the pivot centre

Force Fx= 1N
Force Fy= 1N
Force Fz= 1N
Mx = 1Nmm
My = 1Nmm
Mz = 1Nmm

12688 pm
~0 pm
~0 pm
22111 prad
~0 rad
~0 rad

~0 pm
12688 pm
~0 pm
~0 rad
22111 prad
~0 rad

~0 pm
~0 pm
2043 pm
~0 rad
~0 rad
4.606 rad

The analysis of the ideal pivot provided good results, that is maximum deformation in radial direction was
acceptable. With FE analyses the machining tolerances were evaluated also to define the error budget of
Optical Path piston error. In Table 24 there is the evaluation of the piston error contribution at 0.13 rad:
note that the max value budget is 1.2 pm.
The results of the tolerance analysis are the same for all the considered materials for the reason described
before (the disturbance to be minimize is caused by an elastic load). In the end, three materials were

60 / 126

evaluated, as shown in Table 25. The Aluminium alloy was chosen as baseline because of the weight, the
high actuator force margin and the high torsional pivot frequency. It should be added that the mirror to be
mounted in PAAM is made from Zerodur glass and the Aluminium has the worst thermal conductivity with
respect to the Zerodur. It is possible that the material chosen will be different if the thermal stability will
become more important than the maximum allowable weight.
Table 24: results of the nominal tolerance analysis

Max value
Ideal pivot piston error
Case 1B Blade Thickness: nominal 5 m
Case 2 Blade Length: nominal 10 m
Case 3 Blade Width: nominal 10 m
Youngs Modulus variation: nominal 2%
Mirror mounting tolerance (detail design)
Optical bench alignment tolerance static offset
Direct Sum:

Statistical sum of errors (i )

0.057 pm
0.9 pm
0.0255 pm
0.112 pm
0.3 pm
1.39 pm
0.92 pm

Actually the CSEM is working on this project. A first physical model based on this study will be
manufactured and tested to verify the performances and the stability.
Table 25: comparison between different materials


Copper Beryllium
CuBe C17200 TM04
(BW 190 HM)

(DIN 3.7165,
AMS 4911)

3.58 E-05
89.565 E-03

Al 6061-T6
1.16 E-05
46.119 E-03

Mass (pivot + frame structure) (x10^-3 kg)

Kt (angular stiffness of the pivot) - Nm/rad
K (linear stiffness of pivot) N/m
Inertia kg*mm^2
M (torque to have the max angular
displacement +412 rad) - Nm
F (force on the lever to have the max angular
displacement) - N
F*a (Nom actuator force) N [Cosmos FEM]
F*a (Worst case actuator force) - N
Actuator force margin
Disturbance force
(in the direction of the lever) - N
Torsional pivot frequency (Hz)
Linear pivot frequency (Hz)




10.5 E-03

5.407 E-03

8.619 E-03




1.90 E-05
73.523 E-03

5.5 Solution to decrease the effects of the driving rod

Mr. Henein has suggested an interesting way to decrease the disturbance caused by the transmission
blade. The idea is to place in series two pivots with the same rotation axis one inside the other. The
61 / 126

connection between the two is the heart of the idea: since the transmission blade acts on the intermediate
link, the motion is transferred to the internal link (where there is the mirror) by means of two flexible
blades. Due to the placement of the two blades, the rotation is transmitted to the internal link but the
disturbance load is filtered. The disturbance will be present on the internal link but with a lower order of
magnitude. This idea was not implemented on the PAAM mechanism because of the complexity of the
manufacturing. In fact it is very difficult, almost impossible, to obtain monolithically the double pivot with
redundant lever too.

Fig. 127: 3D view of the doubled pivot

internal link

Fig. 128: the two pivots, one inside the other

5.6 Different solution for the pivot

Fig. 129: different solution for the PAAM main pivot

During the PAAM development a different pivot was evaluated. In this case, the blades in the pivot have
different bonds and different load condition with respect to the pivot actually implemented in the LISA
PAAM. To evaluate the angular stiffness of the pivot, it is possible to use the scheme in Table 26. Note that
in this case the blade has only a different spatial placement. However during flexion, the blades get shorter:
for this reason, if the central link is linked to the fixed link with more than two blades, the pivot is over
constrained (for the same pivot with 2 blades see paragraph 2.3). In this case it was not possible to have
only two blades because of the required radial stiffness: additionally the radial stiffness can be changed
62 / 126

with a different blade number. Moreover the manufacturing of this pivot is simpler than the actual pivot;
however in this case it is more difficult to obtain monolithically the device with the lever as well. Note that
the highest performances in EDM actually are obtained with wire EDM.
Table 26: load case for a blade of the different solution

k =

Jy =

8EJ y 3 p 3 p 2
1 +
+ 2 ;

adml 2
bh 3
; adm =
E (2hl + 3hp )

63 / 126

6 The 2 DOFs nanoconverter

6.1 The original nanoconverter

Fig. 130: Heneins nanoconverter (scheme)

Fig. 131: Heneins nanoconverter (photo)

In a recent paper [25] Henein introduces an interesting mechanism, named nanoconverter: the most
important feature is that with this mechanism it is possible to have a very big reduction factor (between 20
and 1000) in a very small volume (Fig. 130 and Fig. 131). In this way it is possible to use commercial
actuator with micrometric accuracy to have a nanometric accuracy. Note that this device has been
patented. The working principle is explained in the following. In Fig. 132 there is a scheme of the
nanoconverter. If the output link of the nanoconverter should be free to move rigidly linked to the
intermediate stage, it would follow the red path (a parabola, see paragraph 2.3). However since it cannot
move in x direction and because the converting blade has an S shape, it is necessary that it recovers the x
displacement following the blue path (another parabola). Since the two parabolas are shifted one with
respect to the other one of a quantity x0, the resulting motion is the black straight line. The reduction factor
is a function of L and x0: i =

6 x0

x 10


output stage displ.




Fig. 132: Henein's nanocaonverter: the idea




int.stage displ.



x 10

Fig. 133: Henein's nanoconverter: the 2 parabolas

64 / 126

This mechanism, with 1 DOF, is applied in the Differential-Phase-Contrast Interferometer setup in the SLS
Synchrotron Radiation TOMCAT Beamline (Zurich).
The dimensioning of the nanoconverter is quite simple, once the reduction factor is chosen. In fact in the
structure it is possible to find one parallel-springs linear bearing, two 4-circular-hinges linear bearings and a
converter blade. The first stage, the input stage, has the task to connect the linear actuator used to move
the device, allowing the transversal parasitic motion of the second stage, that would be prevented if the
output shaft of the linear actuator was directly linked to the intermediate stage. The displacement of this
stage is equal to the very small parasitic displacement of the intermediate stage: for this reason, it is
possible to use a 4-cirular-hinges linear bearing. The 4-cirular-hinges linear bearing presents, in addition, a
very high transversal stiffness (with respect to the parallel-springs linear bearing), feature exploited to
transmit the load from the actuator to the second stage. The second stage, the intermediate stage, has to
guarantee a long displacement, even longer if the lever ratio increases. For this reason, the second stage is
a parallel-springs linear stage, a stage characterized by a bigger displacement than the 4-cirular-hinges
linear stage with the same volume. The converter blade is dimensioned just as the blades of the second
stage, since the shift does not affect appreciably the behaviour. The displacement of the last stage is in
nanometric scale: the 4-cirular-hinges linear stage guarantees a sufficient displacement with good
performances in terms of transversal stiffness.

6.2 The 2 linear DOFs nanoconverter

The idea was to realize a 2 linear DOFs nanoconverter, based on a parallel kinematics, exploiting the
principle of the functioning of the original nanoconverter.

stage 1

output stage


stage 1


fixed stage

flexure C

flexure A

flexure B

flexure D

stage 2

stage 2



Fig. 135: XY flexure stage equivalent model

Fig. 134: XY parallel flexure stage

The first step was to find an idea to realize a planar parallel kinematics with 2 DOFs: it is necessary to have
2 rectilinear DOFs, locking the rotation in the mechanism plane and all the out-of-plane DOFs. The good
kinematics was found in the Awtars PhD thesis [26] and it is shown in Fig. 134. This figure is one of the
starting points for the PhD thesis, but in this case it is sufficient. In fact, the motion stage has 2 translational
65 / 126

DOFs, that it is possible to control with 2 actuators linked to the 2 intermediate stages. Considering only a
two-dimensional space, the flexible device can be translated in an ideal device as shown in Fig. 135. The
problem is that the displacement of both the intermediate stages is a translation but not a straight line and
it is not simple to connect the actuator to them. Awtar solves the problem using the way that is shown in
Fig. 136 and in Fig. 137, that is a connection able to transfer the axial load but compliant in all the other
directions. Acting with the actuator x there will be a translation of the motion stage in x direction, with a
parasitic displacement in y direction: as shown in paragraph 2.3, it is possible to foresee with a good
accuracy the transversal displacement of a parallelogram flexure.
intermediate stage

Fig. 136: : connection between actuator and flexure
stage (1)

Fig. 137: connection between actuator and flexure

stage (2)

The simple idea to obtain the 2 DOFs nanoconverter is the fusion of the 2 systems presented earlier: in
particular a nanoconverter can be linked to each intermediate stage in the XY stage. In addition it is
possible to fuse the nanoconverter output stage with the parallelogram flexures that link each intermediate
stage to the ground (Fig. 138).

Fig. 138: the idea for the 2 linear DOFs nanoconverter

The result is shown in Fig. 139. The simplicity of the original nanoconverter was maintained, with the
possibility to manufacture it monolithically. This device maintains all the advantages of the original one,
such as the possibility to reduce the cost of the scanning operations, with an interesting improvement of

66 / 126

the performances. Moreover, the dimensioning of the whole device does not present any additional

Fig. 139: the 2-DOFs nanoconverter

6.3 The 3 linear DOFs nanoconverter

After the 2 DOFs nanoconverter it is natural to think of a 3 linear DOFs nanoconverter. The useful
kinematics was founded in the Delta Cube robot [126] (Fig. 140), and the application of the nanoconverter
was performed as in the previous case. The result is shown in Fig. 141. In this case it is very difficult to
obtain a monolithic device: in any case the three kinematics chains are equal, making the manufacturing

Fig. 141: the 3-DOFs nanoconverter

Fig. 140: Delta Cube robot

67 / 126

7 MEMS deformable device for optical applications

68 / 126

69 / 126

70 / 126

71 / 126

72 / 126

73 / 126

74 / 126

75 / 126

76 / 126

77 / 126

78 / 126

79 / 126

80 / 126

81 / 126

82 / 126

83 / 126

84 / 126

85 / 126

86 / 126

87 / 126

88 / 126

89 / 126

90 / 126

91 / 126

92 / 126

93 / 126

94 / 126

95 / 126

96 / 126

97 / 126

98 / 126

99 / 126

100 / 126

101 / 126

102 / 126

103 / 126

104 / 126

105 / 126

106 / 126

107 / 126

108 / 126

109 / 126

110 / 126

111 / 126

112 / 126

113 / 126

114 / 126

115 / 126

116 / 126

117 / 126

118 / 126

119 / 126

8 Devices manufactured in plastics

8.1 Laboratory pincer
This works, depicted in an ampler publication [28], was arisen from a request of the Chemistry for
Technologies Laboratory, at the University of Brescia. A grip device was needed to equip an X-Ray
diffractometer: the specimen to be held was a cylinder, 8 mm in diameter, 15 mm in length. A special
purpose grip device was developed taking advantage from flexural hinges because the standard device did
not fit this specimen. The requirements were met with special reference to the necessity of alignment of
the cylinder with the plan in which X-ray calibration is performed. Fig. 142 and Fig. 143 show the interface
with X-Ray diffractometer. The spring can bear up to about 40 g.

Fig. 142: interface with diffractometer (1)

Fig. 143: interface with diffractometer (2)

The most important problems in designing the grip device were related to the weight of the device itself
and to the alignment requirements of the specimen. The weight problem was solved manufacturing the
grip in plastic material. The required alignment was achieved through the design of the grip. To design the
mechanism reference was made to a Mr. Villas work [1], the one piece drafting compass. The main
difference is the manufacturing process: with regard to production numbers, the compasses are made with
injection moulding while the grip device is machined with a conventional machining process. The design of
the grip device wasnt a particular challenge. The cinematic scheme is simple: to study motion, every
flexure hinge is replaced by an ideal revolute joint. In agreement with Lobontiu [16] a pseudo-rigid-body
model with negligible spring stiffness was used. This approximation is due to selected flexure hinges:
symmetric circular flexure hinges with small dimensions were chosen to simplify the manufacturing. The
second reason of this approximation is the following: the hinges work in large displacements and being
made by plastic material the strength in the smallest thick exceeds the yield strength during the first
motion. Consequently the compliant mechanism was studied as a standard mechanism.
With regard to the material choice, the polyamide 6.6 was chosen because of its mechanical properties: it
has a good yield limit in comparison to Youngs module and its easy to machine. It is also promptly
available. Flexure hinges were obtained by standard drilling and milling operations. The device is operated
by a screw with bounded spring (Fig. 144 and Fig. 145).
After assembly, the device was tested and a peculiar phenomenon was detected: after gripping further
screw rotations induce a specimen raise as shown in Fig. 146.

120 / 126

Fig. 146: specimen displacement

Fig. 145: the pincer into the


Fig. 144: photo of the pincer

It was supposed that this phenomenon was due to a not proper positioning of flexure hinges: Fig. 147
shows the forces acting on the hinges in gripping phase. It seems that a slight difference between the real
geometry of the flexure hinges and the schematized hinges as modelled can justify a different disposition of
the real forces with references to the modelled ones. Such difference, as shown in figure, may give reason
of the unexpected vertical resulting component. After gripping the specimen, the grip is in the position
shown in Fig. 147. The same figure evidences loads acting on link ABC, as derived from pseudo-rigid-body
model (the link A-D is a simple crank, so it is axially loaded). It is clear that the flexure hinge A is the
problematic point: the flexure hinge is almost orthogonal to the direction of the load that is it is loaded not
only in axial direction axially but it is also loaded by a shearing component.


Fig. 147: disposition of the flexure hinges

A modified device was designed, as shown in Fig. 148. In this case special care was paid to hinges
disposition. This is an important conclusion: during the design of a compliant mechanism, hinges must be
located paying attention both to kinematics and to loads acting lines. This hypothesis will be verified after
manufacturing the modified grip device.

121 / 126


Fig. 148: new disposition of the flexure hinges

8.2 Child highchair

Another design study was performed following a request by an Italian company. The idea was to suggest an
intensive use of plastic flexure hinges in the production of a child highchair. Two solutions were proposed:
The first one (Fig. 149) is composed of a support, that copies the Villas drafting compass design with
respect to the kinematics, only in a different scale, and a seat in which a mechanism orients the footrest in
accord with the inclination of the chair back.

Fig. 149: highchair (version 1)

The second one (Fig. 150) is composed of a seat designed to be monolithically moulded flat and a support
with one linkage with only 1 DOF able to set the high of the highchair in the first part of the displacement
and able to close the highchair in the second part.

122 / 126


Fig. 150: highchair (version 2)

For sure these designs are not directly suitable for the industrialization but they can suggest that this
technology can be improved to be interesting for macro scale applications.

123 / 126

The goal of this PhD thesis was to increase and to organize the knowledge in flexure field at the University
of Brescia. Both the synthesis process and the analysis process were faced and the results are reported in
this thesis.
Regarding the synthesis, the most important step was my experience with Mr. Simon Henein and the
researchers of the Systems Engineering Division of the CSEM. Their vision of the design process in
compliant mechanisms based on flexure articulations was entirely adopted. One of the consequences is
that it is not advisable to use the famous term flexure hinge: in fact hinge is associated to an ideal
revolute joint, with 1 DOF. It was shown that the so-called flexure hinges have usually more than 1 DOF.
The use of the term flexure articulations should create less confusion.
The analysis process is shown by means of the study of different applications. In general, the first step is the
evaluation of the behaviour of every blade in the device, considering separately each blade with proper
bonds and loads. In this way it is possible to perform a preliminary dimensioning of the device. The FEM
software has to be used in a subsequential phase, to verify the hypothesis introduced in the first
dimensioning and to evaluate the most complex load cases.
In agreement with the vision of the Mechanical Draw Group of the University of Brescia, where I did my
PhD, a practical approach was adopted: for this reason a lot of applications were presented, showing in
each case the exploited features of the mechanisms based on flexure articulations. When possible, the
devices have been manufactured and tests were performed in order to verify the used models in the
designing phase.
This thesis can also be read as a useful collection of different exempla of mechanisms based on flexure
articulations. Also the naiveties, due to the initial inexperience, are shown and discussed by the light of the
knowledge developed during the PhD: it can be instructive for the beginner, since it clarifies the advantages
of a correct approach to the mechanisms with flexure articulations.
This thesis shows the actual interest, the ample prospects in the compliant mechanism field and the
possibility to apply it in very different domains, from the MEMS (see chapter 7) to the meso devices (see
chapters 3, 4 and 5).

124 / 126

1. Cambiaghi D., Poletti G., Mechanical design of a UHV facility to launch and recover a low-speed
projectile tested on board KC 135, 3rd European Space Mechanisms and Tribology Symposium,
Madrid, ESA SP-279, Dec. 1987.
2. Villa V, Villa G, Villa L; From the concept to the shop, design of a drawing tool: A compass in
polymer; AED 2004 - 4th International Conference on Advanced Engineering Design; Glasgow;
Scotland (UK); September, 5-8; 2004
3. L. Howell, Compliant mechanisms, Wiley, New York, 2001
4. Camploni C., Progettazione e prototipazione di meccanismi cedevoli, Master thesis, Universit
degli studi di Roma Tor Vergata, Anno accademico 2002-2003
5. S. Henein, P. Spanoudakis, S. Droz, L. I. Myklebust, E. Onillon, Flexure pivot for aerospace
mechanisms, 10th European Space Mechanisms and Tribology Symposium, San Sebastian (Spain),
24-26 September 2003
6. Physik Instrumente website:
7. Dynamic Structures & Materials website:
8. P. Spanoudakis, L. Zago, O. Chtelat, R. Gentsch, F. Mato Mira, Extremely high-resolution tip-tiltpiston mirror mechanism for the VLT-NAOS field selector, SPIE Symposium on Astronomical
Telescopes & Instrumentation 2000, 4007-48, Munich (Germany), March 2000
9. E. Hungerbhler, H. R. Burkhard, Device for reducing the force in a force-measuring apparatus, in
particular in a scale. United State Patent 5,340,951 (1994)
10. R. Le Letty , F. Barillot, N. Lhermet, F. Claeyssen, M. Yorck, J. Gavira Izquierdo, H. Arends, The
scanning mechanism for Rosetta / Midas from an engineering model to the flight model, 8th
European Space Mechanisms and Tribology Symposium, Liege (2001)
11. P. Pham, Y.-J. Regamey, M. Fracheboud, R. Clavel, Orion miniangle: a flexure based, double-tilting
parallel kinematics for ultra-high precision applications requiring high angles of rotation, 36th
International Symposium on Robotics(ISR) Proceedings, Tokyo, 2005.
12. AGIETRON Micro-Nano: lavanzata in movimenti di elettroerosione finissimi, AGIE experience, No
26, September 2005, (GF AgieCharmilles website:
13. Overstolz T., Tunable optical microsystems featuring vertical electrostatic comb drives, PhD thesis
- Institute of Microtechnology -University of Neuchtel (Switzerland), 2007
14. Tupperware Italia S.p.a. website:
15. Brigham Young University Compliant Mechanisms Research Group: or

125 / 126

16. N. Lobontiu, Compliant mechanisms : design of flexure hinges, CRC Press, Boca Raton, 2003
17. S. Henein, Conception des guidages flexibles, Presses polytechniques et universitaires romandes,
Lausanne, 2001.
18. Seydel T., Kolln K., Krasnov A., Diddens I., Hauptmann N., Helms G., Ogurreck M., Kang S. G., Koza
M. M. and Muller M., Macromolecules 40 (4), 1035-1042 (2007)
19. Davies R. J., Zafeiropoulos N. E., Schneider K., Burghammer M., Riekel C. and Kotek J. C., Colloid
Polym. Sci. 282, 854866 (2004)
20. Dassa L., Disegno per la miniaturizzazione: sviluppo di un dispositivo di trazione per misure
diffrattometriche 2d di stress residuo, Master thesis, University of Brescia, (2004)
21. A.A. V.V., Aluminum and Aluminum Alloys, ASM Specialty Handbook, ASM INTERNATIONAL, 1993
22. Wang L., Han G. and Zhang Y., Carbohydrate Polymers 69 (2), 391-397 (2007)
23. Dencheva N., Denchev Z., Jovita Oliveira M. and Funari S. S., Journal of Applied Polymer Science
103, 22422252 (2007)
24. D. Gazzoli, Disegno e sviluppo di un giunto universale innovativo a cerniere flessionali, Bachelor
thesis, University of Brescia, (2006)
25. S. Henein, M. Stampanoni, U. Frommherz, M. Riina, The Nanoconverter: a novel flexure-based
mechanism to convert microns into nanometers, Proceedings of the 7th euspen International
Conference, Bremen, May 2007
26. S. Awtar, Synthesis and Analysis of Parallel Kinematic XY Flexure Mechanisms, PhD thesis,
Massachusetts Institute of Technology, Cambridge (USA), February, 2004
27. J.-P. Bacher, Conception de robots de trs haute prcision articulations flexibles: interaction
dynamique-commandes, Phd thesis, cole Polytechnique Fdrale de Lausanne (EPFL), Lausanne,
28. L. Dassa, D. Amadori, Particular types of flexure hinges development and application, 5th
Conference on Advanced Engineering Design, Prague, June 2006

126 / 126

Grazie a tutti gli amici che mi hanno sopportato in questi anni.

Merci tous les amis qui mont accompagn tout au long de ces annes.
Thanks to all my friends who supported me over these years.

Grazie ai mie fratelli, Marco e Giovanni, e ai miei genitori, Cecilia e Battista.

Grazie. Merci. Thanks.

Luca Dassa