annual report - fmtc.be report 2005.pdfgnosis chines 79% 1% t s osts 1% 66% 33% ome gion me y....

2005Annual Report

Annual Report I 2005

003Contents Preface ......................................9

1 INTRODUCTION TO FMTC ......11

1.1 What is FMTC? ......................... 11

1.2 Activities .................................. 11

1.2.1 Modular machines .................. 11

1.2.2 High-dynamic machines ........ 11

1.2.3 Machine diagnosis .................. 11

1.3 Activity chart ...........................13

1.4 Interaction of member companies with FMTC ................ 14

2 2005, FMTC IN GRAPHS ......... 15

3 FMTC’S DEVELOPMENT ......... 17

4 SHORT DESCRIPTION OF THE RESEARCH PROJECTS ....... 19

4.1 Modular machines ................. 19

4.1.1 Performance of field busses .. 19

4.1.2 Motion synchronization ........ 19

4.1.3 Open real-time control framework ...................................21

4.1.4 Industrial adoption of the open real-time control framework ......................................22

4.1.5 Wireless control networks ..... 23

4.1.6 Mobile sensors .......................24

4.2 High-dynamic machines ....... 25

4.2.1 High-dynamic motion control 25

4.2.2 Vibro-acoustic modelling ......26

4.2.3 Active structural vibration control ................................... 27

4.2.4 Micropositioning ...................29

4.3 Machine diagnosis .................30

4.3.1 Teleservice bridge software ...30

4.3.2 Signal processing for diagnosis of machines ...................31

4.3.3 Model-based diagnosis methodology ............................. 32

4.3.4 Machine-integrated torque sensing ................................ 33

4.3.5 Sensor-fusion for torque measurements ........................34

4.3.6 Measurements of temperature distributions .................. 35

4.3.7 Temperature measurements at contact surfaces ........... 36

5 FMTC PUBLICATIONS IN 2005 37

6 FMTC MEMBERS ...................39

7 CONTACT INFORMATION ......41

7.1 Address ................................... 41

7.2 How to reach us ..................... 41

7.2.1 By car (from Brussels or Liège via E40) ............................ 41

7.2.2 By car (from Hasselt or Aachen via E314) ......................... 41

7.2.3 By public transport ................ 41

List of figures .........................43

Annual Report

nPreface

When the Flanders’ Mechatronics Tech-

nology Centre was conceived in 2003, it

was a challenging vision based on col-

lective research between the major me-

chatronic companies in Flanders. Thanks

to the continuous investment of the

member companies, the large support

of the strategic partners (Agoria, WTCM

and K.U.Leuven), the guidance of our

board of directors, and the dedication of

the FMTC personnel, this vision has ma-

terialised in just two years in a research

infrastructure, a highly qualified team,

and a portfolio of research projects. I

would like to take the opportunity to ac-

knowledge all the persons that contrib-

uted to these accomplishments.

The first concrete results of the FMTC

research projects are presented in this

annual report. And with its solid base,

FMTC is ready for more in 2007. I wish

you a pleasant read and invite you for a

fruitful collaboration.

Marc Engels

General Manager

Marc Engels - General Manager


005

2005


006Annual Report I 2005

1: Introduction to FMTC

n 1.1 WHAT IS FMTC?Flanders’ Mechatronics Technology Cen-

tre vzw (FMTC) is a joint research centre

whose mission is to establish the bridge

between the academic and industrial

know-how in mechatronics. The me-

chatronic design approach combines

mechanics, electronics, and software in

an optimal way to make products more

modular, more intelligent, and perform

better. FMTC was founded by AGORIA,

the Belgian federation of the technol-

ogy industry, and leading mechatronic

companies in Flanders. FMTC’s measure

of success is the innovative incorpora-

tion of its research results in members’

products. As a consequence, this will of-

fer the FMTC members the opportunity

to improve the competitiveness of their

products in the global market.

At the end of 2005 FMTC had a total of 17

member companies: Atlas Copco, Barco,

Bekaert, CNH, Daikin, EADS S&DE, Gilbos,

Hansen Transmissions, Teleservice Sys-

tems, LVD, Michel Van De Wiele, Packo

Inox, Pattyn Packing Lines, Picanol, Spic-

er Off-Highway, Televic and IPSO-LSG.

n 1.2 ACTIVITIESTo achieve its mission, the centre con-

ducts industry-driven, mid-term, and

long-term research projects in the fol-

lowing domains:

> 1.2.1 Modular machines

The increasing demand for customi-

zation, scalability and reusability of

present-day machines can only be met

with novel, modular machine architec-

tures. Research is therefore conducted

on two important issues in this modu-

larity concept:

• Wired and wireless communication

techniques between the different

machine modules. This allows the re-

trieval of a maximum amount of infor-

mation from the machines and, at the

same time, optimizes the communica-

tion capabilities between the machine

modules.

• The design of flexible and modular

software architectures for real-time

control with a high degree of reusabil-

ity.

n 007



> 1.2.2 High-dynamic machines

In the future, machines with higher ef-

ficiency, higher speeds and accelera-

tion, higher power throughput and, at

the same time, lower noise emissions

will become increasingly important. The

realization of these, often conflicting

demands, requires new machine design

methodologies and an improvement of

the state of the art in machine control.

FMTC focuses on model-based motion

control and novel (active and semi-ac-

tive) vibro-acoustic control techniques.

> 1.2.3 Machine diagnosis

Machine builders are continuously in-

creasing the performance of their ma-

chines. This entails a growing complexity

which, in turn, complicates the diagno-

sis process necessary to solve problems.

In addition, the economic impact of

system failures is significantly rising as

well, for both vendors and customers. As

a consequence, the need for diagnostic

tools is high.

FMTC therefore focuses on three corner-

stones of machine diagnosis:

• Measurements using intelligent sen-

sors and sensor networks

• Interpretation of the measurements

using signal processing and sensor-fu-

sion

• Exchange of data by techniques such

as advanced teleservice solutions

n 1.3 ACTIVITY CHARTThe research activities and related research projects carried out by FMTC in 2005 are

summarized in Figure 1. The progress and results for each project are described more

in detail in Section 4.

Figure 1: FMTC research activities

009

n Wireless control networks

n Mobile sensors

n Motion synchronization

n Open real-time control framework

n Industrial adoption of open real-time

control framework

n Performance of fieldbus systems

Modular machinesModular machines

n High-dynamic motion control

n Active structural vibration control

n Vibro-acoustic modelling

n Micropositioning

High-dynamics machinesHigh-dynamics machines

Machine diagnosisMachine diagnosis

n Machine integrated torque

sensing

n Sensor fusion for torque

measurements

n Measurements of temperature

distributions

n Temperature measurements at

contact surfaces

n Teleservice bridge software

n Signal processing for diagnosis

of machines

n Model-based diagnosis

methodology

FMTC research activities

Annual Report I 2005 Annual Report I 2005

n 1.4 INTERACTION OF MEMBER COMPANIES WITH FMTCThe core business of FMTC is the defi-

nition, management, and execution

of user-driven mechatronic research

projects that serve several companies.

FMTC offers a team of highly-skilled re-

searchers for performing the research

projects, and has a number of strategic

alliances with academic centres, creat-

ing a professional environment to share

the results. At least three member com-

panies need to be interested in a topic

before a joint research project can be

initiated. By sharing the cost of research,

the investment of the individual com-

panies can be largely reduced. As well

as the joint research projects, FMTC also

conducts contract research for individu-

al companies (who may or not be mem-

ber companies) where results would not

be detrimental to current partners and

projects. FMTC also actively seeks par-

ticipation in European projects.

010 0112: 2005, FMTC in graphs

n

Illustrated in Figures 2 to 5 is a graphical representation of various aspects of FMTC’s

operations.

n

3%

52%

45%

Basic ResearchCollective ResearchContract Research

Figure 2: FMTC resources (man days) by project type

40%34%

26%

Figure 3: FMTC resources (man days) by activity type

Modular MachinesMachine DiagnosisHigh-dynamic Machines

79%

21%

Personnel costOther Costs

Figure 4: FMTC Operating Costs

1%

66%

33%

Figure 5: FMTC Operating Income

Subsidies from the Flemish RegionOther IncomeRevenue from Industry


0123: FMTC’s development

nn

Flanders’ Mechatronics Technology Cen-

tre was founded by AGORIA on 27 May

2003 and began operating on the first of

October 2003 with three employees. In

2004 FMTC experienced a rapid increase

in personnel, a trend which continued

in 2005. At the beginning of 2005, FMTC

employed 9 full time highly-educated

engineers, a number that increased to

14 by the end of the year. In addition,

2 Ph.D. researchers at the department

of mechanical engineering (at the uni-

versity of K.U.Leuven) who worked on

FMTC research formed part of the team.

A photo of the current team is shown in

Figure 6.

In 2005 of the 17 member companies, 7

placed individual employees at FMTC to

facilitate knowledge transfer to member

companies, aiding industrial implemen-

tation of results from the collaborative

research. In total the 7 companies spent

15 man-months in this direct knowledge

transfer.

The major share of FMTC activities in

the past year of operation consisted of

8 projects of basic research (SBO) and

9 projects of collective research (GCO).

An overview of these projects is listed

in Table 1. Notice that 3 of the ongo-

ing projects were completed at the end

of 2005. In addition, a new project was

started in 2005. The fact that 14 of the 17

member companies have taken actively

part in these projects and that 62 user

group meetings were organized in 2005,

indicate that these projects were ex-

ecuted in very close collaboration with

the industry. In addition to the long-

term collaboration agreements, there

was a far-reaching collaboration with

K.U.Leuven-PMA, WTCM and Flanders’

Drive Engineering Centre in 2005. Fur-

thermore, in 2005 there was cooperation

on specific projects with K.U.Leuven-

ESAT and KHBO.

013

Figure 7: FMTC team

PROJECT NAME Project Effective End User- type Start date groups 2005

Performance of field busses GCO 1/04/2004 31/07/2005 3

Motion synchronization GCO 1/01/2004 31/03/2006 4

Open real-time control framework SBO 1/11/2003 31/10/2005 3

Industrial adoption of open real-time control framework GCO 1/03/2004 30/09/2006 3

Wireless control networks GCO 1/10/2003 30/04/2006 6

Mobile sensors SBO 1/10/2003 31/12/2006 6

High-dynamic motion control SBO 13/10/2003 12/10/2007 4

Vibro-acoustic modelling GCO 1/06/2004 31/12/2006 2

Active structural vibration control SBO 1/04/2004 31/12/2006 2

Micropositioning SBO 1/10/2003 30/9/2005 0

Teleservice bridge software GCO 1/10/2003 31/03/2006 5

Signal processing for diagnosis of machines GCO 1/03/2004 31/12/2006 3

Model-based diagnosis methodology SBO 10/11/2003 31/2/2007 3

Machine integrated torque sensing GCO 1/10/2003 31/12/2005 3

Sensor fusion for torque measurements SBO 1/10/2003 31/12/2006 3

Measurements of temperature distributions GCO 1/03/2004 30/09/2006 6

Temperature measurements at contact surfaces SBO 1/01/2005 30/9/2006 6

Table 1: Ongoing FMTC-projects

Figure 6: FMTC’s laboratory infrastructure


014 0154: Short description of the research projects

nThe projects are grouped in three domains:

modular machines, high-dynamic ma-

chines and machine diagnosis (as illus-

trated in Figure 1).

n 4.1 MODULAR MACHINES> 4.1.1 Performance of field busses

Fieldbusses play an increasingly im-

portant role for the realization of com-

plex and automated machines. A large

number of fieldbus systems are cur-

rently available on the market. However,

no single system appears to be, or is

likely to become, an industry standard.

Rather, each individual supplier is trying

to push its system as a standard. De-

ciding if the use of a fieldbus would be

advantageous, and if so, which fieldbus

should be used for a given application, is

a non-trivial exercise. In order to clarify

this situation, FMTC did a case study for

one of the partners and compared (for a

given application) the advantages and

disadvantages of using fieldbusses and

compared the performance and limita-

tions for different systems.

More recently, Ethernet has been gain-

ing popularity for factory automation

and even for use within machines. The

substantial gain in bandwidth compared

to more traditional fieldbusses is the

driving force behind this trend. However,

at the beginning of the project, it was

not really clear if Ethernet is also com-

petitive when it comes to electromag-

netic interference robustness. A number

of tests revealed that, although Ethernet

is indeed more susceptible to electro-

magnetic interference, compared to the

Controller Area Network, it can be used

in harsh electromagnetic environments,

provided that basic rules regarding

grounding and shielding are adhered too.

The project finally made an in-depth

technical comparison of two major Eth-

ernet-based fieldbus systems; Powerlink

and EtherCAT. An implementation of

the latter on top of the eCos embed-

ded operating system was delivered.

Negotiations with the EtherCAT Technol-

ogy Group have started to release this

software under the open source license

framework.

> 4.1.2 Motion synchronization

In many machines, motion profiles of

different moving parts are not inde-

pendent from each other. Traditionally,

the coupling between these different

motion profiles is realized mechanically

(cams, gears, mechanisms). Although a

mechatronic approach towards motion

synchronization will often require more

installed power in the system, there are

a number of inherent advantages over a

pure mechanical solution:

• Flexibility: changing the relation be-

tween different motions only requires

software updates and no longer a me-

chanical redesign of (parts of) the ma-

chine;

• Modularity: if a secondary motion can

be considered to be a feature of a ma-

chine, a mechanical approach neverthe-

less requires the design of two different

machines, one with the mechanical cou-

pling and one without. A mechatronic

coupling on the other hand allows the

feature to be fitted as an optional mod-

ule on an existing machine; and

• Functionality: a mechanical transmis-

sion is almost invariantly character-

ized by high stiffness for every position

and over a broad frequency range. Me-

chatronic couplings on the other hand

allow the designer to influence this be-

haviour if required. An example could be

the isolation of parasitic oscillations to

just one part of the machine.

The FMTC project “Motion Synchroniza-

tion” focused on the third, functionality

advantage, at the same time tried to

realize advanced, adaptive mechatronic

transmissions based on low cost, mod-

erate bandwidth and asynchronous

fieldbusses like CAN.

The use of fieldbusses like CAN for this

kind of applications is only possible if

the required amount of information

that needs to be exchanged between

the participants to the synchronization

process can be limited to the strict mini-

mum, requiring the incorporation of a

priori information on the working of the

machine in every individual motion con-

troller. If it is known beforehand that a

machine is characterized by deviations

at its nominal rotational speed, and that

these deviations can be characterized,

this knowledge reduces the amount of

information that needs to be exchanged,

compared to the case where a constant

rotational speed is supposed and devia-

tions are considered errors.


An important prerequisite for achieving

this is accurate time synchronization

between the different participants to

the synchronization process. An exist-

ing algorithm has been improved and

time synchronization errors below +/-

400 nanoseconds have been achieved.

The algorithm has been demonstrated

with the Controller Area Network, but

can equally well be applied to any other

asynchronous fieldbus.

The possibility of influencing the accu-

racy of the motion synchronization in

function of the position has been dem-

onstrated with a number of laboratory

experiments.

All project results have been consoli-

dated in a number of well documented

MATLAB tools.

> 4.1.3 Open real-time control

framework

European machine tool builders are

typically small and medium sized en-

terprises (SMEs), active in very special-

ized niche markets. Although software

development is not the core business

of these companies, software, and in

particular hard-real time control soft-

016 017ware plays an ever increasing role for the

realization of the added value and spe-

cialized functionality of their products.

Although commercial control solutions

can be bought on the market, these are

so called “on-size-fits-all” solutions that

do not really address the specialized

needs of SME machine builders. To make

things even worse, commercial control

solutions are typically closed source

products which prohibit modifications

to the control software.

This situation is the reason why many

machine tool builders develop their

control solutions in-house. Robotics re-

searchers at the department of mechan-

ical engineering of K.U.Leuven started

the “Open RObot COntrol Software”

project (OROCOS, http://www.orocos.

org) to overcome similar problems with

commercial robot controllers. OROCOS

recognized that all robot controllers

share a great deal of (real-time) common

functionality and that there is great po-

tential for the joint development of a

toolkit or framework for Robotic control

applications. OROCOS is no robot con-

troller by itself but rather offers all the

necessary building blocks for building

one, thereby allowing the researcher or

robot manufacturer to focus on his ap-

plication or research instead of difficult

and error-prone embedded and real-time

programming. In order to push the pos-

sibilities of joint development to its lim-

its, and at the same time safeguard the

openness of the OROCOS framework, an

open source development approach and

an open source licensing scheme were

chosen.

FMTC has actively supported the ORO-

COS initiative, and the underlying PhD

work, from the beginning of the project.

Great effort has been spent in broaden-

ing the scope of OROCOS, from a mere

robotics control framework into a much

more generic real-time control frame-

work that is also useful for building

machine tool controllers and contains

the fundamental building blocks for any

other advanced real-time control appli-

cation.

During the last year, OROCOS has been

extensively tested on a number of in-

dustrial applications from FMTC part-

ners and several reference papers have

been produced. More or less in parallel

the accompanying PhD thesis has been

written and is, at the moment of writ-

ing, awaiting its defence.


018 019> 4.1.4 Industrial adoption of the open

real-time control framework

The OROCOS real-time control frame-

work is the result of four years of PhD

research. More emphasis has been put

on generating as complete and correct a

framework as possible, than on deliver-

ing a production grade software prod-

uct. OROCOS may be called complete

in a sense that it contains all the funda-

mental building blocks for building any

advanced real-time control application.

This, unfortunately, makes it also very

complex and not really accessible to the

average control engineer, who does not

always have much experience in em-

bedded and real-time programming. To

make things even worse, the outcome of

the OROCOS project is only available as

source code and depends on a number

of other open-source initiatives which

are still evolving and often also only

available as source code. In conclusion,

the project resulted in a very powerful

framework, but with limited use for the

average control engineer. However, there

exists significant demand to expand the

practical implementation.

FMTC recognized that, in order for ORO-

COS to be used in industry, two major

prerequisites needed to be fulfilled:

1. Ease of use and ease of installation

needed to be drastically improved. ORO-

COS needed not only to be accessible to

professional embedded and real-time

hackers with a strong interest in control,

but also, and even more important, to

the regular control engineer with an in-

terest in software; and

2. Stability of the code base needed to

be improved, or maybe better, guaran-

teed. Whereas the stability and quality

of the code was already very good, and

certainly more than sufficient for re-

search purposes and doing robotic ex-

periments, industrial users need almost

certitude that the software they put in

their control systems will run for many

years without errors.

In order to address these issues FMTC

has, during the last year, packaged the

OROCOS framework and all other soft-

ware it depends on, thereby greatly

simplifying the installation process. In-

stead of manually verifying that version

dependencies between the different

components are honoured and building

each individual component from sourc-

es, installation has become as simple as

updating a standard Linux distribution

with a few binary packages. The packag-

ing process itself has been automated to

the maximum extent possible in order

to keep this activity maintainable in the

long run.

Not only has the packaging process

been automated, but also the testing of

the software. Packages are automatical-

ly (regression) tested after building and

only if they pass these tests, released.

A second, major goal of this project

is the industrial adoption of the open

source development philosophy. In order

for this to happen, an active user com-

munity needs to be established around

OROCOS. Usability on more than just

the GNU-Linux and RTAI operating sys-

tems on which OROCOS has been devel-

oped, could be a catalyst for such a user

community to be formed. FMTC has for

this reason also started porting OROCOS

to the eCos operating system.

> 4.1.5 Wireless control networks

As a consequence of the increased mod-

ularity of machines, the need for com-

munication is mounting. Wireless com-

munication is an attractive possibility

to address this need, because it avoids

wiring costs, offers enhanced flexibility,

and supports mobility. However, the in-

dustrial environment poses some chal-

lenges for wireless communication:

multiple reflections, time-variations of

the propagation channel, interference,

dust, etc. Moreover, control applications

pose very strict requirements on laten-

cy and robustness of communication.

Therefore, this project investigates the

applicability of wireless communication

technologies in an industrial environ-

ment.

The project mainly focused on the Blue-

tooth technology and several modules

were evaluated. First, detailed measure-

ments with the Casira module from CSR

(see Figure 8) were performed in order to

have a better understanding of both the

Bluetooth technology and the internal

mechanisms of the Casira module itself.

From the results, it appeared that the

retransmission scheme of Bluetooth is

influenced by a polling mechanism: the

master of the communication polls the

slaves for data at regular intervals. This

polling scheme has a dramatic impact

on the latency of a packet. Indeed, when

a packet is lost, it will be retransmitted


only when this polling action occurs,

thus increasing the roundtrip latency.

Two other modules were also evaluated:

a Bluetooth to CAN gateway from RM-

Michaelides and a RS232 to Bluetooth

gateway from PhoenixContact. Both

showed a similar polling behaviour, re-

sulting in comparable latency and ro-

bustness characteristics.

For large deployments of Bluetooth pi-

conets, the inter-piconet interference

can be the limiting factor. To evaluate

this interference behaviour, a dedicated

simulation tool was developed. The tool

contains the Bluetooth hop selection

kernel and requires a detailed definition

of the deployment geometry and also of

the propagation conditions. As a conse-

quence, propagation measurements had

to be performed to obtain the necessary

inputs. The results showed that machine

covers can become crucial to make large

Bluetooth deployments feasible.

Standard Bluetooth modules do not

allow controlling low-level communi-

cation parameters. However, some la-

tency-critical applications will only be

feasible with an optimized setting for

020

these parameters. Therefore, the next

step in the project will be the develop-

ment of a Bluetooth control profile that

will offer this flexibility to the users with

a minimum of required effort.

> 4.1.6 Mobile sensors

In many industrial machines it is attrac-

tive to mount sensors on moving parts.

However, such a situation yields an im-

portant challenge, both for the commu-

nication with the sensor and the power

supply to drive the sensor. The current

solutions are expensive and are sensi-

tive to wear. Therefore, we investigate in

this project a wireless solution for both

aspects.

Figure 8: Casira module from CSR

n 4.2 HIGH-DYNAMIC MACHINES> 4.2.1 High-dynamic motion control

This project aims at the development of

a design procedure to identify a param-

eter-varying dynamic model of a ma-

chine and to design a high-performance

varying (motion) controller based on an

identified system model. The main chal-

lenge in this procedure consists of guar-

anteeing the stability and performance

of the scheduled controllers.

In 2005 methods have been calculated

which can be used for the identification

of Linear Parameter Varying systems

(LPV) and for the systematic design of

scheduled controllers. The focus was

thus on the identification of LPV sys-

tems rather than on the analytic LPV

control design for LPV systems. These

methods were tested on a laboratory

setup, being the Flex-cell, an industrial

pick-and-place machine that was rede-

signed in 2004.

Although many interpolation methods

have been successfully applied by control

engineers, these methods are either ad-

hoc and require a lot of trial and error, or

they fail when the order of the system is

021In 2005, we continued with the investi-

gation of inductive power transmission

to moving parts. The main conclusions

were:

1. Intermittent inductive power trans-

mission is attractive for moving ob-

jects;

2. Radio communication is preferable for

wireless data transfer; and

3. Local pre-processing of the sensor sig-

nal is required for several applications.

Consequently, a plan was developed to

realize a mobile sensor node based on

these assumptions and cooperation

with IMEC was established.


too high. The methods developed within

this project offer a systematic approach

that can handle complex systems.

First a method has been developed to

interpolate between several linear sys-

tems that are identified in fixed operat-

ing points. This technique starts from

an affine interpolation between the

poles, zeros and gain of the local sys-

tems as a function of the varying pa-

rameter and results in a global varying

system description with a simple affine

state-space representation. The method

is successfully applied on the Flex-cell,

not only for identification purposes, but

also to a more systematically designed

gain-scheduled controller. Experimental

results show the benefit of the proposed

method.

However, as the resulting affine state-

space representation is not minimal,

the outcome of analytic LPV control

methods would turn out to be conserva-

tive. To prevent this, the interpolation

method was refined to come up with a

minimal affine parameter dependent

model of the system. This is done by

fitting special type of functions on the

variation of the poles, zeros and gain

of the fixed models. Also this method

is successfully applied on the Flex-cell

and the resulting affine models can be

used to derive analytical LPV controllers.

The synthesis of these controllers is cur-

rently under examination.

In the near future the performance of

the analytically derived LPV controllers

will be compared to the scheduled con-

trollers designed using the developed

interpolation method. Also a criterion

will be worked out to decide if ad-hoc

gain-scheduling controllers can be used

or if the more advanced LPV controllers

should be used.

Next to the motion control design for

the FlexCell, also the motion control of

off-road vehicles was further worked

on. Off-road systems primarily show

discrete varying dynamics, related to

the transmission ratio the drive train is

working in. As such, these systems are

traditionally controlled using feedfor-

ward methods. Special attention in this

research was not only paid to pure feed-

back control of the varying dynamics,

but also to the effects of a transition be-

tween feedforward control and feedback

control strategies.

To improve the quality of shifting be-

tween gears, the slip over the clutches

was measured and used as a feedback

signal. Using the slip signal, a better

trade-off could be made between the

driver’s perception and the transmis-

sion’s life time, which has been shown

experimentally.

To date, measurements have only been

completed on a test bench. In the future

measurements are planned on an off-

road vehicle.

> 4.2.2 Vibro-acoustic modelling

This project aims at investigating and

developing easy-to-use methods with a

low degree of complexity to evaluate the

acoustic behaviour of a newly developed

machine in an early stage of the design.

More specifically, this will be done by

relating structural measurement results

of a machine to acoustic measurement

results of the same machine. These

methods will provide the designer with

a tool to improve the acoustic behaviour

of a new machine design.

As it was decided at the end of 2004

that the methodology to be developed

should be implemented on a generic

022 023test setup, in the beginning of 2005

an inquiry was made to determine the

components that the industrial partners

would like to see in the test setup and

the measurement technologies they are

interested in.

The inquiry resulted in a conceptual test

setup with a gearbox, comprising three

gears, six bearings and a housing. The

setup was further detailed and designed

so that the preload torque on the gear-

box can be set equal to the nominal

torque of the gearbox, which is impor-

tant to study its noise emission as an

industrially representative case. As the

gearbox is preloaded mechanically us-

ing a toothed belt, the size of the driving

motor, and its disturbing noise, could be

reduced significantly.

In the next step, in 2006, a dedicated

acoustic shielding will be made for each

of the main components of the setup

(motor, belt and gearbox). Then, struc-

tural and acoustic measurements will

be carried out on the setup. Interpreting

and relating these measurement results

will allow identifying the setup vibro-

acoustically. Based on results, some

structural changes will be made on the


setup and the measurement and analy-

sis procedure will be repeated. Compar-

ing the results of the initial setup to the

results of the modified setup will finally

allow determining the vibro-acoustic

impact of the structural changes made

on the setup. This knowledge can then

be used to improve the vibro-acoustic

behaviour of a new machine design.

> 4.2.3 Active structural vibration control

Traditionally, active control of planar

structural radiators has approached the

problem through integrating sensors

and actuators on the structure. In such

an approach, the goal is to modify the

systems response to disturbance, rather

than influence the disturbance path to

the radiating structure. Based on a de-

tailed partner inquiry many practical

issues, which ultimately limit the effec-

tiveness, of such an approach appeared.

A general problem shared by our part-

ners is noise radiation from a structure

housing a rotating device. In this project

we have attacked the problem by seek-

ing to reduce the force transmission in

the transmission path, which is the root

cause of the sound radiation, rather

than the vibration on the radiating sur-

face itself. It is hoped that higher levels

024 025of attenuation may result.

In the last year (2005) the design of a

passive setup, which was initiated in

2004, has been detailed. The setup fur-

thermore has been built and tested (il-

lustrated in Figure 9 and Figure 10). The

aim of this stage of the project was to

create the platform of the active ele-

ment. The passive setup was capable

of noticeable noise levels, replicating

the problems faced in industry. The pas-

sive design was deliberately designed

to simulate a wide range of industrial

problems. This was achieved by using

a modular design which allows for dif-

ferent methods of excitation, variation

in rotation speed, different unbalanced

loads and different planar radiators. The

design of the active bearing element has

been completed and the manufacturing

of the element is in its final stages. Il-

lustrated in Figure 11 is a sketch of the

active element. An important part of

the active element design was that it

be straightforward to integrate into the

housing surrounding the shafts of many

of our partners. Ideally it should be an

independent unit which replaces current

bearings and takes up a limited amount

of room.

Based on an initial literature survey of

actuator/sensor pairs for active struc-

tural damping, piezo-electric technol-

ogy was chosen to be used for the ac-

tive set-up. A dynamic model of the

setup was constructed and based on

the results of several dynamic simula-

tions a commercially available piezo ac-

tuator/sensor pair was selected. In first

instance this actuator/sensor pair will

be placed around the bearing to inves-

tigate the possibilities of ‘in the path’

structural vibration and noise control.

Shortly initial tests will begin where the

integration of the actuator/sensor into

the active element, level of control au-

thority and different control laws will be

examined.

Figure 9: Sketch of the passive setup, motor on the right

Figure 10: Photo of the passive setup

Next to the possibilities for active con-

trol with active sensor/actuator pairs

around bearings, the applicability of

such elements between different parts

of the housing or on the radiating plate

of the setup will be investigated in a par-

allel track.

The ongoing review of published litera-

ture and contact with industry indicates

that this work is both novel and has the

potential to result in a practical solution

for a wide range of industry.

Figure 11: Sketch of the active bearing element.


026> 4.2.4 Micropositioning

In this project, we wanted to achieve

three objectives:

1. Design and testing of a piezoelectric

motor with a speed that exceeds 100

mm/s and with a traction force of 1N

and a position resolution of 10 nm;

2. Modelling of the piezoelectric motor

described above, to be able to gain

more insight in the possibilities and

limitations of piezoelectric motors in

general; and

3. To understand whether this piezoelec-

tric motor or piezoactuators in general

are suited for application in industrial

machines.

In 2005 we finalized this project, with

the following results:

1. The maximum speed measured on the

built piezomotor (see Figure 12) was

about 550 mm/s and its maximum

Figure 12: Built piezoelectric motor

traction force was 0.45 N. Another

prototype motor was built achieving

a traction force of 1 N. The position

resolution is down to the nanometres

level, determined by the piezoactua-

tor resolution.

Based on the constitutive equations of

piezoelectric materials, a lumped-pa-

rameter model for a piezoelement was

derived, including its rate-independent

losses. This model is easy to integrate

in mechanical systems comprising pi-

ezoelectric actuators as well as piezo-

electric sensors. (This model was applied

in the project of active structural vibra-

tion control.) A complete experimentally

verified motor-model was developed to

investigate the effect of resonance of

the motor-performance and on the heat

dissipation of the piezoactuators.

In 2005, a PhD thesis, entitled “Develop-

ment of fast, stiff and high-resolution

piezoelectric motors with integrated

bearing-driving function”, was written,

incorporating the research carried out

on this topic between 2000 and 2005.

Also two journal papers have been sub-

mitted for publication.

027n 4.3 MACHINE DIAGNOSIS> 4.3.1 Teleservice bridge software

Although much of the advanced func-

tionality of present machines and

machine tools would not have been

possible without mechatronics, the ac-

companying increase in complexity of

machines confronts end-users with an

increasingly difficult maintenance prob-

lem. Such, complexities create opportu-

nities for machine builders to take over

(part of) the maintenance responsibility

from their customers.

New business models are however, inevi-

tably, accompanied with new challenges

for which a technological answer needs

to be found. An advantageous situation

for both end-user and machine builder

can only exist if the latter can exploit the

advantages of scale. In order for this to

happen, the machine builder needs to

know the exact status of the installed

machine at the customers location.

Although commercial applications for

tackling this telecommunications prob-

lem exist on the market today, solutions

are far from generic and typically require

tedious and difficult integration efforts

in order to connect customers’ ma-

chines with the manufacturer’s service

centre. It is exactly this generality that is

the subject of the FMTC research, which

aims at a generic communication frame-

work for connecting any mechatronic

device with any service centre. At the

same time, care was taken to come up

with a scalable solution that is not only

suitable for static but also dynamic (ma-

chine) configurations.

Generic interfaces that make abstraction

of protocols, devices or service centres

have been defined. Translation between

these generic interfaces and device or

service centre specific interfaces goes

through so-called gateway components;

dynamically bound, and isolated pieces

of software with a mere protocol trans-

lation responsibility.

Platform independence was a very im-

portant requirement to the bridge soft-

ware. JAVA has therefore been chosen as

the development language for the bridge

software and CORBA for describing in-

terfaces and as middleware technology

for integrating gateway components.

In the course of 2005, the bridge soft-

ware has been finalized and, for demon-

strating purposes, gateway components

have been implemented for the Com-


mon Industrial Protocol (EthernetIP)

and towards a proprietary service centre

provided by one of the FMTC member

companies.

> 4.3.2 Signal processing for diagnosis

of machines

The goal of this project is to develop

methodologies for monitoring and diag-

nosing mechatronic systems. In 2004 it

was decided to focus the project towards

signal processing and to investigate sig-

nal processing aspects of sensor-reduc-

tion, sensor-fusion and sensor-reduc-

tion. Since it was concluded that these

techniques could be seen in a broader

framework of (multisensor) data fusion,

in 2005 the objectives for the remainder

of the project were specified and detailed

in a general datafusion framework.

The established global objective of the

project is to provide a data fusion frame

to combine data from different sensors

and features/parameters to obtain more

accurate inferences and identification of

system anomalies.

The specific objectives are:

• To provide a usable generic frame of

signal processing techniques for (mul-

028tisensor) datafusion and diagnose;

• To demonstrate signal processing tech-

niques on a concrete problem present

at several member companies, namely

the detection and identification of

damage at bearings and rotating parts;

and

• To produce a well documented tool-

box for the developed tools and algo-

rithms.

After defining three cases where the

signal processing techniques are going

to be applied, an extensive literature

overview has been made about vari-

ous techniques available for detecting

bearing damage. Based on this study

and the specific damage cases present

at the member companies participat-

ing in the project, it was decided to use

three different types of sensors to cover

a frequency range starting from 10 Hz to

100 kHz for detecting bearing damage.

An accelerometer, a shock pulse sensor

and acoustic emission (AE) sensor were

selected.

A mobile acquisition setup was cre-

ated and measurements were made on

weaving machines with good and bad

camshaft bearings and on a camshaft

test setup. Tests on a third case, a gear-

box with a defect bearing, have been

planned for 2006.

Within the framework of feature selec-

tion from demodulated spectra (the

amplitude variation of the sensors’ sig-

nals at its resonant frequencies), several

enveloping methods were implemented.

In addition, a study about feature sub-

selection and feature extraction, which

is necessary to reduce the dimension of

the feature space for classification has

been carried out. In 2006 more meth-

ods for feature selection will be imple-

mented. Furthermore, feature reduction

methods will be combined with classi-

fiers to detect and identify the bearing

defects from the three test cases.

> 4.3.3 Model-based diagnosis

methodology

Having access to machines at the cus-

tomer’s premises is one aspect of re-

mote maintenance and diagnostics,

being able to interpret measurements

and machine data and draw conclusions

with respect to machine health from

them, is another.

Machine diagnosis is nowadays still

029performed by human experts, making

(service) companies critically depend-

ent on key service personnel. To make

things even worse, expert knowledge

is seldom available (or even known) in

an explicit form, making the training

and education of service technicians a

difficult and time consuming (mostly

experience based) activity. The lack of

explicit expert knowledge makes it also

very hard to automate many aspects of

machine diagnosis.

On the other hand, many processes and

machines (or parts thereof) can be mod-

elled, meaning that a huge amount of

information and knowledge is available.

In addition system models normally

contain, although also in an implicit

way, the same knowledge on system di-

agnosis as a human expert.

The aim of the Model-based diagnosis

methodology is to gain insight into how

system models, and which kind of mod-

els, could help to solve, or at least reduce,

the above mentioned problems. For this

reason, and after doing a thorough lit-

erature review, a case study has been se-

lected at one of the FMTC member com-

panies. A detailed physical model for the


selected machine has been derived and

is currently used for simulating the ma-

chine’s behaviour in both nominal and

faulty conditions. These simulations will

in a next step be used for the derivation

of threshold values on signals (or com-

binations of signals) that are a strong

indication for faulty behaviour of the

machine. In other words, to gain insight

in where and what to watch in order to

be able to diagnose the machine. Mak-

ing use of models and simulations has

a number of important advantages over

pure experimental work:

• A priori knowledge of the machine’s

working and construction can be used

in systematic way;

• Simulation can be a lot faster than

setting up a lot of individual experi-

ments;

• Simulation allows to do experiments

(and learn from them) that would de-

stroy a real machine;

• Simulation gives much more control

over the boundary conditions that in-

fluence machine behaviour; hence al-

lows isolating the machine’s working

from the environment; and

• Simulation provides a history of every

parameter that is modelled. This would

require a huge amount of sensors on a

real setup and even allows monitoring

parameters that can not be measured

on a real machine or for which a meas-

urement would interfere with the ma-

chine’s operation.

In the future, estimators will be imple-

mented that will allow the online moni-

toring of the selected relevant parame-

ters and the conclusions with respect to

diagnosis of the selected system will be

validated experimentally.

> 4.3.4 Machine-integrated torque

sensing

The goal of this project is to develop a

low-cost torque sensor that can be in-

tegrated in the mechanical structure,

based on existing technology. At the

start of the project (in 2003), a survey on

state-of-the-art torque sensing princi-

ples was made. Based on this survey four

technical principles for torque measure-

ment were selected and analyzed in a

static test set-up. As from this analysis

it was concluded that the polarized band

method was a promising technology, it

was decided to further focus on that

technique. A set-up that allows experi-

mental verification of the dynamic be-

haviour (speed and torque) was built for

030that purpose. Based on the results of an

extensive measurement campaign the

polarized band method was accepted to

be a possible candidate for integration

in the applications of our partners.

Therefore, in 2005 a detailed specifica-

tion of the applications of the partners

in the project was put together and the

applicability of the polarized band meth-

od for these applications was discussed

with MDI, who manufactures sensors

based on this technology.

Figure 13: Weaving machine installed at the facilities of FMTC to test an integrated torque sensor.

A specific application, being the torque

measurement on the main axis of a

weaving machine, was selected as a test

case for the design and manufacturing

of an integrated torque sensor.

A weaving machine has been installed

at FMTC (see Figure 13) and specific meas-

urements (vibration levels, magnetic

field strengths, etc.) were performed to

detail the environmental conditions of

the prototype sensor to be designed. The

design and manufacturing of the sensor

itself was completed by MDI.

031


032After manufacturing, the prototype

(see Figure 14) was shipped to FMTC for

detailed testing. In order to have a refer-

ence value for the torque signal, strain

gauges were installed on the shaft of the

prototype sensor.

After integration, the prototype has been

thoroughly tested on properties such as

linearity, hysteresis, accuracy, magnetic

influences, etc. This experimental test-

ing of the prototype is currently in a fi-

nal stage and the sensor is expected to

perform as specified.

To be able to assist our members in pur-

chasing the sensor in large quantities, a

business discussion with MDI has been

initiated.

Figure 14: Prototype of the torque sensor to be integrated into the weaving machine

> 4.3.5 Sensor-fusion for torque meas-

urements

The goal of this project is to exploit the

torque estimation by multiple torque

sensors to increase the overall accuracy

of the torque measurement at an attrac-

tive cost. This is justified by the fact that

retrieving additional torque data from a

setup can often been easily done (e.g.

through reaction forces of a motor and

drive parameters in a frequency-con-

verter).

In 2005, the initial plan that was made in

2004 for this project has been detailed

as resource became available towards

the end of 2005. Two torque estimation

principles were worked on in more de-

tail: torque estimation of an induction

motor using current and velocity meas-

urements and torque estimation by

measuring reaction forces. On the first

topic FMTC collaborated with ELECTA,

a division of the Electrical Engineering

Department of the K.U.Leuven and a

torque estimation algorithm was devel-

oped and implemented on a set-up in

the lab of FMTC. This set-up consists of

two induction motors coupled by a shaft

and a reference HBM torque sensor. The

performance of this algorithm was ana-

lyzed both experimentally and in simu-

lation. Analysis of torque estimation by

measuring reaction forces was initiated

in 2005. Next to the further analysis and

development of these torque estima-

tion principles first steps were taken in

the transfer path analysis of rotational

drive lines and sensor fusion techniques

such as Kalman filtering. These topics

will be further elaborated on in 2006 and

experimentally analyzed on a new (to be

designed) rotational test set-up.

4.3.6 Measurements of temperature

distributions

The goal of this project is to measure

temperature distributions. Contactless

measurement of temperature distribu-

tions are especially of interest in labora-

tory experiments for the optimisation of

a design. Various applications were en-

countered where infra-red cameras were

not suited and hence other technologies

are required.

In cooperation with group-T, a Master

thesis was conducted on temperature

measurements with temperature sensi-

tive paints. This thesis was successfully

finalized, with the realisation of a test

set-up and the demonstration of two

types of reversible paints (paints which

change colour repeatable based on their

temperature).

In parallel the project investigated an al-

ternative approach to measure tempera-

ture distributions combining data from

a limited number of discrete tempera-

ture sensors with a real-time thermal

model of the set-up. However, it quite

rapidly became clear that for most of

the considered applications an accurate

and real-time thermal model could not

be obtained within a reasonable time.

Therefore the project was completely

focused on the direct measurement of

temperature distributions with tem-

perature sensitive paints. In a first step,

these paints were characterized in more

detail. Aspects like accuracy, hysteresis,

illumination conditions, camera gain,

Figure 15: Paint response time set-up

033


etc. were investigated. Further on, the

dynamical behaviour of the paints was

measured. To this end, the set-up illus-

trated in Figure 15, was adapted.

4.3.6 4.3.7 Temperature measurements

at contact surfaces

For several applications, there is no di-

rect visibility of the temperature meas-

urement location. As such existing

techniques, like infra-red thermography,

and upcoming techniques, like tempera-

ture sensitive paints, are not applicable.

Moreover the wiring of discrete temper-

ature sensors, especially on moving ob-

jects, can be cumbersome. To this end,

this project investigates the feasibility

of alternative technologies.

In cooperation with group-T, a Masters

thesis was defined and initiated. The

purpose was to study the possibilities of

fibre optics in measuring the tempera-

ture on locations without direct optical

path. Up till now, an overview of various

fibre optics temperature measurement

techniques has been made. A combina-

tion of optical fibre with temperature

sensitive paints was selected for this

project. The test set-up to evaluate this

technique was also defined.

034 0355: FMTC publications in 2005n

• B. Koninckx and P. Soetens, “Open source real-time machinesturingen”, Techniline,

01 April 2005.

• P. Coenen and B. Koninckx, “Softwareontwikkeling voor embedded systemen in

auto-industrie en machinebouw”, Techniline, 13 April 2005.

• P. Soetens and H. Bruyninckx, “Realtime Hybrid Task-Based Control for Robots and

Machine Tools”, Proceedings of the International Conference on Robotics and Auto-

mation 2005 (ICRA’05), Barcelona, Spain, 18-22 April 2005.

• A. Van Vlierberghe, “’Condition monitoring’ van de oliekwaliteit”, Techniline, 13 May

2005.

• W. Van de Vijver, S. Devos, H. Van Brussel and D. Reynaerts, “Piezoelectric Linear

Drives”, 2nd International Workshop on Piezoelectric Materials and Applications in

Actuators , Paderborn, Germany, 22-25 May 2005.

• W. Symens and I. Hostens, “Actieve geluids- en trillingsdemping”, Techniline, 15 July

2005.

• R. Belien, “Een nieuwe machinerichtlijn?”, Techniline, 21 October 2005.

• B. Koninckx, S. Eeckhoudt and P. Lamsens, “Bewegingssynchronisatie: beschikbare

technologieën en nieuwe ontwikkelingen”, Techniline, 2 December 2005.

• B. Koninckx and L. Somers, “Teleservice: naadloze connectie tussen machine en

diensten’’, Techniline, 9 December 2005.

• S. Devos and S. Eeckhoudt, “Piëzo-elektrische transformatoren: compact en ge-

schikt voor automatisering”, Techniline, 9 December 2005.

• W. Symens and S. Masselis, “Low-cost koppelmeting in standaard machine-uit-

rusting”, Techniline, 23 December 2005.


036 0376: FMTC Membersn

n

At the end of 2005 FMTC, 17 companies have joined FMTC: Atlas Copco; Barco;

Bekaert; CNH; Daikin; EADS S&DE; Gilbos; Hansen Transmissions; Teleservice Sys-

tems; LVD; Michel Van De Wiele; Packo Inox; Pattyn Packing Lines; Picanol; Spicer

Off-Highway (Dana Corporation); Televic and IPSO-LSG. Illustrated in Figure 16 are the

member company logos.

Figure 16: Logos of the FMTC member companies


0387: Contact info

n 7.1 ADDRESSFMTC

Campus Arenberg, Celestijnenlaan 300D

B-3001 Heverlee, Tel.: +32 16 32 80 50,

Fax: +32 16 32 80 64, www.fmtc.be,

[email protected]

n 7.2 HOW TO REACH US7.2.1 By car (from Brussels or Liège via

E40)

Leave the E40-highway Brussels-Leuven-

Liège (Luttich)-Aachen at the exit “Leu-

ven - E314”. You have to leave the E314

(A2) highway right at the exit “Leuven”

(exit number 15). Along the highway con-

nection E40/E314-Leuven, turn right at

the second set of traffic lights. You are

now in the Celestijnenlaan. The FMTC

building is 500 meters further down this

street at your righthand side. You enter

via the entrance in the middle (see Figure

17). The reception of FMTC is at the third

floor.

7.2.2 By car (from Hasselt or Aachen via

E314)

You leave the E314 (A2) highway at the

exit “Leuven” (exit number 15). Along the

nhighway connection E40/E314-Leuven,

turn right at the second set of traffic

lights. You are now in the Celestijnen-

laan. The FMTC building is 500 meters

further down this street at your right-

hand side. You enter via the main en-

trance. You enter via the entrance in the

middle (see Figure 17). The reception of

FMTC is at the third floor.

7.2.3 By public transport

Leuven is along the railway lines Brus-

sels-Liège-Köln/Welkenraedt/Eupen

and Brussels-Liège/Hasselt/Genk. It also

has direct connections to Antwerpen-

Mechelen and Aarschot-Hasselt-Tonger-

en. The Leuven railway station is located

on the right side of the map (see Figure

17). To reach PMA you may either take a

taxi or a bus (line 2 direction Campus).

This bus takes you via “Leuven Centrum”,

“Heverlee Station” and “Kantineplein” to

the Arenberg II-Campus. You get off at

the “Campus Arenberg II”-bus stop. FMTC

is located right across the road. You en-

ter via the entrance in the middle (see

Figure 17). The reception of FMTC is at the

third floor.



040

n List of figuresFigure 1: FMTC research activities ........................................................................................................................ 13

Figure 2: FMTC resources (man days) in 2005 by project type ........................................................................... 15

Figure 3: FMTC resources (man days) in 2005 by activity type ........................................................................... 15

Figure 4: Operating Costs in 2005 .......................................................................................................................16

Figure 5: Operating income in 2005 ....................................................................................................................16

Figure 6: FMTC’s laboratory infrastructure ......................................................................................................... 18

Figure 7: FMTC team .............� 17

Figure 8: Casira module from CSR ...................................................................................................................... 23

Figure 9: Sketch of the passive setup, motor on the right ................................................................................ 27

Figure 10: Photo of the passive setup ................................................................................................................. 28

Figure 11: Sketch of the active bearing element. ................................................................................................ 28

Figure 12: Built piezoelectric motor .................................................................................................................... 29

Figure 13: Weaving machine installed at the facilities of FMTC to test an integrated torque sensor. .............33

Figure 14: Prototype of the torque sensor to be integrated into the weaving machine .................................. 34

Figure 15: Paint response time set-up ................................................................................................................ 35

Figure 16: Logos of the FMTC member companies ............................................................................................. 39

Figure 17: Map for reaching FMTC .......................................................................................................................42

annual report - fmtc.be report 2005.pdfgnosis chines 79% 1% t s osts 1% 66% 33% ome gion me y....

Documents