CS WAVE1: Virtual Reality for Welders’ Training

L. Da Dalto†, D.Steib††, D. Mellet-d’Huart†††, O. Balet†

† CS, Virtual Reality Department †† AFPA, DSI ††† AFPA

5, rue Brindejonc des Moulinais,BP 15872, 35 Boulevard Jodino,BP 211 DI Virtual Reality Forecasting Unit
31506 Toulouse Cedex 5, France 69 632 Vénissieux Cedex, France 93108 Montreuil Cedex, France
Tel +33 (0)5 61 17 64 96 Tel +33 (0) 4 72 90 21 89 Tel : +33 (0)66 312 5556
[email protected], [email protected] [email protected] [email protected]

Abstract
Virtual reality (VR) is a set of technologies that one often confines to R&D projects or high
technology applications for innovative sectors. Nevertheless, tangible interfaces and VR,
when specified from end-user's needs, can also lead to relevant and accepted systems for both
the learning and training in the most common and traditional industries. This position paper
presents CS WAVE, one of the first systems in this field applied to the learning of sensory-
motor skills for welding. It results from an R&D project funded by the European Commission
in the frame of its 5th framework programme. During two years, CS and AFPA have
specified, developed and assessed the WAVE system in close collaboration with end-users. It
is henceforth an inescapable product that has proved efficient for the learning of the welder
profession.

CR Categories and Subject Descriptors: Virtual Reality, Vocational Training, E-Learning,

Virtual Environments for learning, Tangible Interfaces.
Additional Keywords: Welding.

1 INTRODUCTION
In 2002, CS and AFPA have launched the WAVE R&D project with the support of the
European Commission. This project, which successfully ended in 2004, aimed at developing
CS WAVE, the first system using Virtual Reality (VR) technologies to train welders. This
project was coordinated by CS, a European leader in virtual reality application development,
and co-developed with AFPA, the main operator of the French Labour Ministry for vocational
guidance and training of unemployed adults.
The objective of this project was more than the sole development of an innovative system. It
had first to provide an accurate answer to training difficulties in the particular field of
welding. Moreover, the project had to guarantee the acceptance of the proposed system by
manual workers and trainers who can be reticent about new technologies.
Indeed, the introduction of innovative technologies in the general industry is a complex
process. Often starting from a research phase, this process develops through out long periods
of validation and assimilation before progressively integrating the industrial cycle. It is a
question of demonstrating and re-demonstrating the relevance of the proposed solution to the
industrial actors while providing concrete argumentation and figures to the most conservative
of them. If this seems rather feasible for engineering industries, it is far more complex when
dealing with more traditional occupations such as those employing manual skilled workers.

We will see in the following sections that our strategy was not to replace current training
practices but to enhance the educational support proposed to the trainees during the first
phases of the training programme course. At this stage, novice trainees use huge amounts of

1
CS WAVE is a trademark of CS and AFPA, http://wave.c-s.fr for more information.
expensive welding consumables while learning by doing and practicing to acquire sensory-
motor basic skills of welding. Our aim was also to provide trainers with means to closely
monitor and assess the gesture of the trainees.
In addition to the development tasks, we have demonstrated in this project that VR
technologies are efficient for the trainees to support the acquisition of sensory-motor skills to
master their body position and movements. Moreover, they lower the cost of the training
courses. Finally, the architecture of our system enables the seamless integration of CS WAVE
within the existing training process, without disrupting the organization, and while improving
the quality of the learning and the educational dialogue. This was definitely the second key of
success of the project.

2 RELATED WORK
VR technology is increasingly used to support learning [BUR2003] [LOU2003]
[WIN2002a][MAN2001]. As shown in previous experiments, it can be efficiently used to
learn technical procedures [RIC2000], medicine [RIV2003], military strategy [RIC2002],
occupational skills [LOF1995] or academic contents [MOS2002], etc. But amongst all
previous experiments, very few applications reached their final public as the Hubble project
[LOF1995], the Virtual facet [DRO1997] and VIVART [DCR2005] did. As far as we may
know, none of that have been commercialised. As described later on, CS WAVE intended to
be used in real training situation on a large scale. It have been commercialized since March
2004.
Regarding the design of virtual environments for training, former approaches often referred to
idiosyncratic concepts of virtual reality (e.g. immersion, presence, realism) to set up learning
situations. Therefore, expensive immersive interfaces were used, photorealistic visual
feedback and credible sounds were required. Some systems might also have used haptic
feedback in order to recreate the realistic sensations of manipulating tools. Burkhardt & al.
[BUR2003] proposed an alternative to the “realism-oriented” approach. It is based on multi-
sensory structures to afford the cognitive and learning activity of the trainees. This approach
is grounded on one side on empirical investigations of learning in VR, which suggest that
additional sensory information enhance the learning of abstract concepts [WIN2002b]. Winn
claims that new references to sciences as biology of cognition are needed [WIN2003].
Therefore, it is grounded on studies of action, perception and cognition [NOE2000], which
have emphasized the strong relationships between senses, such as between vision and touch.
Referring to work realized by neuroscientists, Mellet-d'Huart [MEL2004] proposes a "model
of (en)action" in order to support counter-intuitive design methods for virtual environments
for learning. He shows, for instance, how conceptualization and simplification of a scene may
facilitate the understanding of a process, and how accurate guidance may facilitate the
realization of embodied and embedded activities. Besides a conceptual frame of references, he
proposes mechanisms that orient the design of the virtual environment depending on the
learning objectives, the characteristics of the learning content and the activity of the learner.
Regarding virtual environments dedicated to train welder few other projects exist [FAS2004]
[HAS2005]. They are still experimental projects and integrates merely industrial or well-
focused educational issues.

3 TRAINING WELDERS
The skilled welder knows, with little visibility, how to decrypt information from the welding
pool to produce an high quality welding. The trainee, on his side, does not have any clue to
understand what he is expected to do and how they could correct their action from visual
hints. Up to now, he had to keep on practicing over numerous tries to finally acquire accurate
sensory-motor skills.

3.1 A skilled motion combined with a permanent concentration

Electric arc initiates the melting of an added metal with the object to join. It is the welding
pool. Either a generator or a welding machine provides electric energy that is required to
weld. This energy supply is preset on the machine by the welder or modified during the
welding. The most used processes are the GMAW one (Gaz Metal Arc Welding) and the
SMAW (Shielded Metal Arc Welding). The welder uses a specific tool to provide the
additional metal: through the torch for the GMAW (a wire running out of the torch) or
enclosed in electrodes that have often to be replaced for the SMAW. Elements to join can be
flat or curved (usually tubular) and form plan or angles.
The welder wears a helmet with a dark glass preventing him from the intense light produced
by the electric arc. A thick jacket and gloves are also necessary to avoid burnings due to metal
ejections and to the heating of the object to join.
The welder, while working, observe the welding pool and modify the position of his tool or
the setting of the welding machine to maintain a welding suitable to the strict specifications.
The welder must master welding speed, trajectory of his motion and distance between the end
of his tool and the object to welding. At the same time, he has to control the welding pool and
interprete what he perceives. Quality of his work may be controlled after cleaning the
welding, visually of by a radiography. Sometimes, a bad welding may be repaired but most of
the time his work must be destroyed, even for a single error. Today, skilled welders are very
demanded as there is a great deficit of them in the industry.

3.2 Difficult, long and expensive learning

The learning of welding is both hard for the trainee and the trainer.
Helmet and gloves limit the trainee mobility as soon as he starts learning in a real situation.
Helmet window shortens the field of view and, with the electric arc, he can only see a small
area of the zone where the metal is added. At the beginning, trainee does not know how to
analyse the modifications or the causes of the modification of the welding pool. With these
constraints, he needs to try to hold and move his tool in a known direction but with no
guidance. Moreover, noises, heat, and sparklings disturb him. He may be afraid of been
burned.
A beginner does not have any idea of how to evaluate a speed, mesured in centimeters by
second and how to maintain it. In the same way, he does not imagine how to keep a distance
of some millimeters between the MAG nozzle or the electrode and the sheet to join. It is even
more difficult when the motion is not linear. These parameters are invisibles. Thereby a
beginner cannot perceive the visual, audio or sensitive clues that guide the skilled welder are
not reachable for a beginner. Transfers of kinematics and kinetics knowledge from a position
to another (from vertical to horizontal for example) is either partial or impossible. This
implies a further training.
The trainee builds to himself a representation of a good motion by practicing with repeated
demonstration, observing motions and the feedback provided by the trainer. He must acquire
one by one all positionning and technologic elements for each welding situation. Most of the
time, he needs to compare what he see or what the trainer says concerning a welding cord, to
pictures memorized of this same cord in course.
So, he practices a lot until his work may be accepted by his trainer (by afterward analysis or
radiography). Weeks and weeks of practice are often necessary before achieving any success.
A lot of metal sheet goes to the trash. Trainees may often feel discouraged and depressed.
Some of them give up and drop out..
For the trainer side, practical welding training is tricky. His training is mainly based on his
own know-how. Any pedagogy of exploration finds its limitations by the little that can be
perceived by a beginner and also by costs that limit the tries. Trainer must guide trainees to
the efficient work by letting them discover as much as possible. Practical demonstration is the
main solution. If the trainer observes the trainee while is welding, he can not see his posture
(due to his helmet) and if he wants to observe his posture, he will not be able to see his work.
So this learning takes a long time.
At AFPA, training courses of GMAG and SMAG start by flat joints during 21 weeks. A full
training takes 6 to 8 months depending on the process and includes tube welding and complex
works done by a team. Trainees spend around 5h30 per day in the workshop where they
prepare and realise their practical training. Welding itself takes 35% of this time. 1h30 hours
is dedicated to technology, drawing and plans reading.
Around 60 AFPA centres train several thousands trainees each year in France thanks to a
hundred trainers. Welding are also a skill integrated in others trainings (maintenance,
mechanics, …). While the result is not correct, the trainees do and redo their welding work.
Sheets and consumables are lost and are very expensive. As it is a long training, this learning
costs a lot. The cost reduction is then a permanent issue.

4 EXPLORING A SOLUTION : THE WAVE PROJECT

Until the arrival of welding robots, there have been no expectation to improve the practical
learning of the welding. Thus, robots technologies, then simulators, did not bring the expected
solution. Later, virtual reality appeared. In 27 months and with the support of European
Commission, a multidisciplinary team had been developing the future CS WAVE.

4.1 Solving learning difficulties

It was necessary to break down the learning process to make it easier. The acquisition of the
sensory-motor skills had to be isolated in regard of its own complexity. It was also important
to provide the trainee with guidance and help him to correct his posture and sensory-motor
activity while limiting the trainer interventions. Thereby, trainees will become more
independent and proactive in his practical training, with more cost efficiency.

4.2 The Virtual Reality perspectives

In a first time, prospective of solutions to improve the practical training conditions has been
oriented to welding robots application. Computer modelling of welding tool position by the
robot may answer to our needs. Each parameter is modelled separately and a computer
manages their application simultaneously. So, we had here an expert system to compare to the
trainee work in real-time.
Nevertheless, the cost of the integration of such a technology was far too important and the
project has been rejected. But a little welding robot has been manufactured. Trainees may,
then, experiment the influence of the welding machine set-up and the welding speed.
Then, the simulators appeared and improved. They provides breakdown and synthesis of the
training situations and environments. This may have been interesting integrated in welding
robots models. Still, this solution does not allow to turn invisible or hidden information,
visible for the trainee to help his learning.
Then, virtual environments have been studied, even if they were still “Research &
Development” and “prototype” oriented. During 2000 and beginning of 2001, AFPA project
team looked for VE oriented to the training. Some of them was too far from their needs (shop
management, …). Other was dedicated to expensive prototypes and experiments (flight
simulators, …).Learning situations were too closed to the reality. AFPA found again the same
limitations for the welding training in these simulations. VE interface was too sophisticated
(immersive helmet, …) for the trainees and trainers not used to computer tools.
Nevertheless, thanks to this new approach, simplification of professional situations seems
possible. It allows to turn visible events, actions or notions usually invisible or abstract, in
real-time. Simplification of the pedagogical work requests to be fulfilled that the transfer from
simulation to reality was still possible. Trainers will then find an help in these tools rather
than be deprived. Multiplication of VE applications and important cost reductions also
motivate us to continue our studies in this domain.

4.3 Project objectives, funding and partnership

To start the specification of a VE dedicated to the welding trainers in the workshop, AFPA
integrated unavoidable characteristics : Length of sheet to weld, realism of sheet to join and
welding tool realism.
It must be necessary to turn visible what was not (speed variations, distance between tool and
sheet) thanks to visual guidance and real-time feedbacks. Little variation of some millimetres
must be easily sensible.
Technologically, solutions seems to exist at a reasonable cost. Still there was the conception
and development cost. Then CS, expert in integration and marketing of Virtual Reality
applications for the industry, proposes a solution with the help of European funding.

4.4 The EUTIST-AMI project and its results

In 2001, CS and AFPA presented a project for a virtual environment dedicated the welding
motion learning. It has been accepted in the 5th Framework Program of IST – Information
Society Technology – by the European Commission. It has been integrated in a cluster of 17
projects called EUTIST-AMI. Started in 2002, it has been closed in march 2004 with success.
The project WAVE gave birth to the product CS WAVE which has been awarded twice in
2004 (productivity award at INDUSTRY 2004 and research and education award in LAVAL
VIRTUAL).
The following schema shows the way the works has been conducted. It also presents the
organisation of CS and AFPA partnership. CS is in charge of the developments and
manufacturing of all the prototypes. Today, CS manufactures and markets CS WAVE.

Definition and corrections of the set-up for the practical exercises

Computer model development, pedagogical features integration

01/2002
α α β β 03/2004
Creation and hardware design of the workbench

Pedagogical conception and improvement of trainer-product interactivity

Pedagical conception and interaction updates

Figure 1: WAVE project timetable (AFPA actions in green, CS actions in blue)

AFPA provided technical specifications of welding processes and a didactic analysis of the
training activity related to the ‘didactisation” [PAS1999]. Specifications included
representations of usable knowledge, determinations, key values, position and motion of the
tool, motion control, body positioning and working instructions. Learning situation is
organised around the issue to solve : comprehension and mastering of the parameters
controlling the tool positioning and holding. They are studied one by one and then applies
together. This process must be introduce to the trainer thanks to a progression table. The
progression plan has been defined by the trainer by observing real training. Knowledge
acquisition is improved by visual guides indicating the limits of the tolerances for each
parameters in real-time together with a general realistic appearance of the welding cord.
Finally, trainee and/or trainer may ask for a debriefing to follow afterwards or in real-time the
trainee work. Results representation must be provided to analyse each try and the overall
progression.

AFPA organised experimentations of the various prototypes : 50 trainers have been involved.
The results helps CS all the way thought the development.
CS WAVE experimentation starts during the European project with two prototypes that
included nearly all features. 24 trainees and 3 trainers, in an AFPA centre in France, starts
motion training with this new tool. Experimentation included two points : Features
stabilisation and pedagogical improvement. The first one has been finished at the end of the
project. The second one started at the same time, in March 2004.

5 THE FINAL PRODUCT : CS WAVE

The final result of WAVE project is a ‘manufactured’ product as it fulfils all technical,
pedagogical and financial requirements. Moreover, CS WAVE reached its final user after a
long experimental period with real trainees.

5.1 Learning characteristics

All elements leading directly or indirectly to a knowledge didactisation are here. All welding
technical specification are included.

5.1.1 CS WAVE answers to the welding technical specifications

It can replace partially the trainer to provide to trainees exercises similar to those done in the
workshop, with the same level of requirements concerning the motion parameters. Local
updates are also possible, depending on the feedbacks of the users.

5.1.2 Pedagogical characteristics

The trainee is placed in his working environment by introducing the notions to master,
processes for progressing and the parameters related to position and motion of the tools. The
requested body position is still demonstrated on a video while preparing his work, before
using CS WAVE.
Learning situation is organised around the issue to solve : the understanding and control of the
motion parameters through a progressive approach by mastering each parameter one by one
and then associating them. Process is demonstrated to the trainee thanks to a table displaying
the overall progression and his status with the results he has already produced.
Knowledge acquisition and identification by the student is possible both by the real-time
feedbacks during the simulated welding and by the analysis of the results afterwards through
their graphical representation.
A representation in different format of a try and its comparison with the two latest one helps
to understand the progression. The trainer can follow in real-time or afterwards each trainee
activity. Debriefing is possible at any time on the request of the trainee or the trainer.

5.2 Technical characteristics

CS WAVE technical characteristics answer the training requirements as presented previously.
We may see hereafter how they are implemented.

5.2.1 The welding workbench

A welding workbench [Figure 2] has been designed within the project. It features a mobile
TFT screen, which automatically positions according to both the user’s height and the
exercise to perform, an embedded PC with an Ethernet port, and a tangible interface that is a
welding torch or an electrode handle.
The latter are the only means available to the trainee to interact with the workbench. They are
the real tools of the welder and are referred as the “welding tool” in the remaining of this
paper.

Figure 2: the welding workbenches

In order to cope with electro-magnetical interferences and occlusions that inevitably occur
during training session, we needed a new sort of tracking system to instrument the welding
tool. This has been developed by InterSense in close collaboration with the consortium. It is
now marketed by this company as the IS-900 PCTracker, a six degrees of freedom motion
tracking system based on a hybrid technology of inertial and ultrasonic tracking. The position
and orientation of the tracking are determined by the output of the accelerometers and gyros.
Drift correction is accomplished in InterSense’s advanced Kalman filter by fusing the output
of the inertial sensors with range measurements obtained from the ultrasonic components. The
result is a full 6-DOF tracking of the welding tool that is very smooth, precise, and free from
jitter.
In the following, we briefly present the successive stages the trainee has to pass to complete a
training exercise.
5.2.1.1 Identification
The first dialog box requires the user to choose his name, in the list of persons authorized to
use the workbench, and to enter his four-digit password. Once performed, he is identified and
recognized by the workbench which repositions the screen at his height. If connected to the
network, the workbench automatically informs the CS WAVE server and, incidentally, the
control centres that a trainee is about to perform an exercise.

5.2.1.2 Exercise definition

The following dialog boxes graphically present the different exercises and the various options
available to the trainee. The trainee can then successively select the different options of the
exercise with his welding tool [Figure 3].

Figure 3: defining the exercise

Once fully defined, a video presents the movement to be learned and the various gestures to
master.

5.2.1.3 Performance assessment

Each exercise is based on pre-defined parameters that are monitored during the exercise in
order to assess the trainee’s performance. These can be speed, distance between the welding
tool and the steel sheet, angles or complex trajectories to follow.
The performance is assessed taking into account the selected level of difficulty, which defines
the error tolerance for each parameter.
For the lowest level of difficulty, the users can also active visual helpers (i.e. 3D metaphors)
that indicate, for each parameter of the exercise, the quality of his performance and the respect
of the constraints.

5.2.1.4 Exercise performance

Then, the workbench first places the screen in the right configuration for horizontal or vertical
welding. Then, a virtual representation of the metal sheets to weld is displayed at real scale on
the screen.
The trainee positions his welding tool at the beginning of the welding cord to create. A set of
visual indicators indicates the ideal origin of the cord. While performing the exercise, the
created welding cord is rendered in three-dimensions [Figure 4] under the welding tool and a
realistic sound is played.
For each parameter of the exercise, a corresponding visual helper, if activated, appears in
green (or disappears) when the trainee stays within the zone of tolerance. It flashes in green
when he approaches constraint limits and becomes red when the user gets out of constrains.
All the monitored parameters are transmitted in real-time to CS WAVE server which
broadcasts them to all the trainees’ control centres. Thus, each trainee can follow any exercise
from any control centre.
 Figure 4: performing the exercise

5.2.1.5 Results
At the end of the exercise, the system provides a graphical representation of the trainee’s
performance. The first information is the level of success of the exercise. If positive, the user
can automatically validate the level corresponding to the exercise.
Then, detailed results are made available in the form of graphs along the welding cord for
each monitored parameter [Figure 5]. The analysis of these graphs, usually with the assistance
of the trainer, will allow the user to better understand his errors and determine how to improve
for the next try.

Figure 5: results

5.2.2 The control centre

The control centre is the software tool used by the trainers to remotely define, monitor and
review exercises [Figure 6]. It can also be used by the system administrator to create and
manage user profiles for trainers and trainees. In addition, the latter can consult their statistic
and progress with the control centre tool.
Control centres can be installed either inside (local area network) or outside (wide area
network) the training centre in order to provide remote expertise and monitor trainees activity
from distant locations.

Figure 6: the control centre interface

If the welding workbench is a totally autonomous system only requiring a power socket and a
network connection to run, the control centre is a software application requiring a Windows
XP/2000 host.

5.2.3 The server

The server is in charge of the network communications between the different elements of the
CS WAVE network. It is also in charge of the persistence and the consistence of distributed
data.
Indeed, the information available within a CS WAVE network is either static (GUI skins and
configurations, exercise parameters, videos) or dynamic (results of exercises and result
syntheses). Both sorts of information are stored in the server’s database and replicated on all
the control centres and workbenches, the role of the server being to maintain the consistence
of these databases.
Installed on a PC dedicated to this usage, the server can be controlled by the system
administrator through a dedicated interface enabling backup, storage and update procedures. It
also provides him with synthetic information on the state of the complete system as well as
networking statistics.

5.3 Integrating issues

As seen in the related work, very few applications have reach their public and are used in
everyday-learning situation. This means that very few work exist that deal with the particular
issues of how to integrate a virtual environment for learning in a vocational training situation
? How the curricula, the pedagogy, the role of the trainer, etc. are affected by the use of this
training resource?

5.3.1 Between observation and statement

5.3.1.1 Usage description

Today, CS WAVE is exploited in 14 training centres of AFPA network. Each centre is
equipped with one or two workbenches. The control centre is installed on the trainer desktop.
These workbenches are installed in a room close to the workshop to avoid dust and noises
perturbation. In June 2006, all of the 60 centres of the AFPA will be equipped.
Training program stays the same. CS WAVE use is controlled by the trainers who allow each
of their trainee on exercise on the workbench depending on his progression. Right from the
beginning of the training, CS WAVE has been introduced to the trainees. Individual and daily
access period are programmed. The training organisation try to use the workbench at a 100 %
rate. Trainees work on CS WAVE each day during ½ or ¾ of an hour. Total duration of the
training is the same. Adaptation period to CS WAVE and its features is necessary. It changes
a little their usual training in the beginning.

5.3.1.2 User statement

Trainees and trainers are usually satisfied by what CS WAVE can bring : concentration and
professional motion. Trainees adapt themselves very quickly to what CS WAVE provides us :
Easy and tangible interface thanks to real welding tool, choices proposed. Moreover they
appreciate the self-evaluation in real-time thanks to visual guide. Visual representation of the
results afterwards focus them also, even if it requires some efforts to understand. They feel
that the integration of this tool improve their training.
Trainers find the same set of exercise existing in the training program. They are happy to be
able to enable/disable them depending on the trainee. They feel confident in creating or
modifying personalised progression. But better than that, real-time or afterwards follow-up
features are very appreciated. They were afraid that CS WAVE slows down their work and
requires more time from them, it is not the case. CS WAVE fits perfectly to the training
program with, at least, the same level of success as before.
The impact of CS WAVE on the sensory-motor skill acquisition is recognised by now. It
reduces the number of trials a trainee has to complete in real situations within the workshop,
to acquire those skills. Moreover, the transfer of the sensory-motor skills training from virtual
to real situations is nearly immediate in the welding cabin. They can then focus themselves on
the electric set-up and their influence on the welding pool.

5.3.1.3 Updates concerning the learning situation

The goal of this project has been to provide an assistance to the self-training of the welding
motion. The need for repeating demonstrations by the trainer and tries by the trainee without
any guarantee for results is one of the main weakness of the current training organisation
based on the trainer.
Splitting the training by isolating the hardest part of the motion learning from the
technological part is efficient. The need for a simplication of the learning process to be able to
cope with its difficulties is possible thanks to its virtualisation. The pedagogical possibilities
offered to the trainer and the trainees to acquire the motion parameters are very wide. The
management of the parameters separately and progressively all together was an important
specification. It is one of the rules of the AFPA concerning the pedagogical process : starting
from the simple, going to the complex.
Training is simplified as it is then possible to see the constant speed, the distance between the
tool and the sheet, the variations in millimetres or degrees. The virtual environment provides a
pedagogical basis to build a representation of invisible and abstracted objects. Self-learning is
possible as its base is included in the modelling.

5.3.1.4 Updates in the trainer practices

Is there really a self-training anyway ? Through observations and exchanges, it appears that it
requests to be assisted by the trainer. It is essential to take times to introduce CS WAVE and
how it works related to the existing training. This must even be repeated at any time.
The trainer needs also to reach the best condition of use of CS WAVE for each trainee, from
the calibration of the screen position, to the time and duration of his training session on the
workbench. CS WAVE requests a high concentration from the user and so the trainer must
observe each trainee and avoid tiredness and discouragement implies by the time spent on the
machine.
This follow-up is simplified by the CS WAVE tools dedicated to the trainer. Exercises history
allows him to follow the progression of a trainee on a single exercise, for each parameter and
so it helps him to decide how to adapt the progression depending on what is wrong or what
has been acquired. Real-time follow-up provides a close diagnostic of the issues. A trainee
may be assisted when he needs to be and on what exactly was wrong by a simple observation
and debriefing. With the same duration, they are now more equipped to assist the trainees :
make them integrate all knowledge in their mental representation and help those who are in
great difficulty.
It is also interesting to observe how the trainees manage themselves their training and
comparing that with their success in the workshop. After some days, they know what CS
WAVE brings them and they can express it very clearly. In the same way, after some training
weeks, they perceive the difference between the real welding (as soon as they feel the welding
pool updates and what happens at the end of the electrode) and the virtual one. They are then
tempted to work only in the workshop whereas CS WAVE training is still necessary for him.
CS WAVE does not reproduce the realism of the welding situation. The trainers are in charge
of the Virtual-Real interface. They learn progressively to exploit the advantages of this VE
and the non-realistic elements transposed in the real welding training. They are still
mandatory to the success of practical training with CS WAVE. CS WAVE should have
include far more functionalities to be able to partly exempt them.

5.3.1.5 Economic impact

It is necessary to quantify the economic impact on a representative number of users.
Nevertheless, we know that the first hypothesis based on a 30 % cut in raw material cost
should be even more important. By analysis what happens at an individual level and even if
the results are not numerous on site using CS WAVE, we can expect that the number of tries
necessary to obtain an acceptable work will be shorten by 50 %. To reach this level, the
Trainee/ Trainer/ CS WAVE system must be pedagogically operational.

5.4 A good integration and optimisation of CS WAVE use : tracks to investigate

5.4.1 Pedagogical scripting of the practical learning with CS WAVE and the workshop.
The progression of the parameters association (15 stages for a level) is a pedagogic variation
of what the trainer is doing in the workshop. Extracted from a multidisciplinary study, it build
a base thanks to a compilation of observations. From now on, this granularity in the approach
of the welding motion learning has never been reached.
To improve and optimise the use of CS WAVE by the trainers, two approaches will be
conducted at the same time. The first one is a pedagogic one where every learning behaviour
will be studied especially if they show a lack of efficiency. The other one is quantitative by
using a questionnaire to retrieve CS WAVE users feedbacks at the end of their training on the
workbench.
These two indicator will help us to better understand how CS WAVE is used for selft-training
and self-evaluation together with what this technology brings to the training.
Moreover, to be able to conduct wide study of learning with and without CS WAVE, potential
partnership with research labs are also studied.
With the current AFPA organisation, the 35 % of welding time are associated with other
practical training : plan study, preparation, cutting, positioning, cleaning, analyse with the
trainer and repeated… after sending the sheet used to the trash. Concentration for welding is
requested on short periods and in between them, the trainee does several other tasks.
CS WAVE captures this concentration immediately and nearly in a permanent way. In a full
use situation during a session, the welding work with CS WAVE may be of 85 % of the full
session time ( 5% navigating, 5% results study, 5 % relaxation). In fact, virtual welding ratio
(cumulated time of welding with CS WAVE by total of individual sessions time) is around 30
%. Nevertheless there are important individual differences. This ratio can goes up to 65%
occasionally and is around 45 % for trainees that validate one or more levels. This confirms
reality or causes : tiredness due to the concentration is one of them. Finally, trainees use
different learning strategies and have specific practice needs for the same efficiency. At that
time, CS WAVE can only be an indication.
This tool is an real innovation. It was not possible to predict its advantages and limits for
trainees and learning. Trainers contribute to define and optimise its pedagogic uses. They are
participating to the validation and updates of the exercises settings and of the pedagogic
hypothesis. With the support project team, they are looking for :
How to preserve trainees motivation by avoiding unnecessary tries,
How to improve the progression between stages or exercises
This may leads to recommendations on how to set-up CS WAVE depending on the training
organisation. But, the study ongoing may also require more important changes.
5.4.2 Economic optimisation
Cutting in the cost of raw material depends on the good use of CS WAVE by the trainers. On
a pedagogical point of view, the use of CS WAVE might become essential. A follow-up by
the AFPA project team and good practice exchanges are planned to help and ensure the
integration of CS WAVE in the trainers processes. On the scheduling aspect, full booking of
the tool during the working time is also a goal to reach. Managers and trainers will be
checking that. The study of the way the trainees are using CS WAVE may also lead to update
scenarios to extend the use of this tool. On an accountant side, there are still efforts requested
to have a better view of the saving realised and then, increase its value.

6 CURRENT STATUS AND FEEDBACKS

CS WAVE has been used by now for nearly four years. Today, it is used by all AFPA training
centres. The deployment plan has started in summer 2005 and will cover all welding training
centre in France for this organisation. More than 100 workbenches will be installed.
At the same time, CS WAVE has been tested and installed in other training organisation in
France and abroad. At the end of 2005, FOREM (Belgium) and UIMM (France) will be
equipped. Other customers from Japan, UK, Spain, Morocco, Scandinavia and Eastern Europe
are interested. A worldwide marketing network is in construction.
Different results have been achieved as an enhancement of the learning process or a dramatic
reduction of the drop out of trainees…

6.1 Improvement of the practical training : less retirement during the beginning of
training
It is an obvious hypothesis knowing CS WAVE advantages. More efficient learning, easily
assimilated should prevent frequent retirements observed today. We must especially study the
efficiency of CS WAVE on the learning issues. This efficiency will be more admitted if it can
maintain more people in the training process.

6.2 CS WAVE and the motion mistake diagnostic

Several tries done with CS WAVE confirm that it can contribute to analyse a motion mistake
for a beginner or even for skilled welders. Issues existing in the practical learning at the
workshop will also occur on CS WAVE. It is then easy to analyse, formalise and treat them
by an exchange between trainee and trainer, when this does not already happens between
trainees.

6.3 Other uses perspectives for CS WAVE

The AFPA will try to use CS WAVE to test motion know-how learned by experience, with all
the requested care, linked to the system limitation and to the distance to the real welding.
In the two forthcoming years, all information associated to CS WAVE use will be formalised
and dispatched. Other improvement are studied like the improvement of the results for the
final certification. Indicators and observing process will be studied, defined and developed.

7 PERSPECTIVES
For a welding training organisation, CS WAVE is a technological and pedagogical
innovation. For the AFPA, the innovation has been in the management of conception and
integration of a Virtual Environment. This project opens new perspectives. Technological
possibilities of the Virtual Reality may answer to various pedagogic issues. Economic context
itself is fair. The cost of this product is not an issue and can even be lower if there is more
demands. Some of its expensive elements (motion tracking system for example) are not
necessary for having any VE running.
We are now looking for new evolutions of CS WAVE like the analysis of a realistic welding
pool or TIG welding on tubular objects.
CS WAVE practical experience, linked to the theoretical evolution of the conception
processes [MEL2004] allow us to improve the effort attached to the software and pedagogical
conception for an efficient specification. Nevertheless a strong collaboration between users,
developers and managers is mandatory on all the project duration to produce good results.
Finally, thanks to CS WAVE, CS and AFPA acquired and improved their skill in managing
and deploying industrial projects based on Virtual Environment dedicated to the training. The
field of application of this kind of projects is wide and is still very open today.

8 CONCLUSION
This paper has presented CS WAVE, a combination of virtual and physical artefacts that form
an efficient and homogeneous structure for the e-learning and training in a traditional field of
industrial activity. CS WAVE was awarded twice. It wan "the excellence in productivity”
prize by a group of 33 journalists from the industrial press at the INDUSTRY 2004 exhibition
in Paris and “the science and education trophee” of Laval-Virtual 2004. Since the end of the
project, we have been approached by numerous international distributors, which promises a
fast exploitation at an international level. It is today commercially exploited by CS, and is
currently deployed within the whole network of AFPA’s training centres (60 centres ).
It is also important to stress that the success and the acceptance of this system could not have
been guaranteed without the strong and constant synergy between CS engineers and AFPA’s
trainers and without the adequate combination of the complementary expertises in Man
Machine Interfaces, Virtual Reality hardware and applications, usability, and e-learning.All
this clearly demonstrates that an innovative system using VR technologies can become, if its
development is driven by real end-user needs, a viable and accepted product that breaks
through very quickly after the R&D effort in the quite conservative industrial world.

9 REFERENCES
[BUR2003] BURKHARDT J.-M., LOURDEAUX D. & MELLET-D'HUART D. La
conception des environnements virtuels pour l'apprentissage In FUCHS P. & MOREAU G.
Eds. Le traité de la réalité virtuelle. Presse de l'école des Mines de Paris, 2003 – in French.
[DCR2005] D'CRUZ M., COBB S., EASTGATE R., KERR S., PATEL H. & WILSON J.
Case studies in the/she application of virtual environments for education and training. In
RICHIR S. & TARAVEL B. EDS. Proceedings VRIC'05, First International VR-Learning
Seminar, Laval, April 20-21, 2005. pp 85-94, Angers, ISTIA, 2005.
[DRO1997] Drouin A., Schmid A. & Thibault G. Formation et réalité virtuelle : résultats de
l'expérimentation dans le stage 7235 "robinetterie : diagnostic conduit de chantier". EDF
DER, 1997. – in French.
[FAS2004] FAST K., GIFFORD T., YANCEY R. 2004 Virtual training for welding.
Proceedings of the third IEEE and ACM International symposium on mixed and augmented
reality (ISMAR 2004)
[HAS2005] HASHIMOTO N., KATO H., IKEHARA R. 2005 MR training system for
manual arc welding using stereoscopic HMD and multi markers. THWAITES H. ED.
Proceedings of thethe eleventh international conference on Virtual Systems and MultiMedia:
Virtual reality at work in the 21st century. Oct. 3-7, Ghehent, pp. 563-571 Budapest,
Archaeolingua.
[LOF1995] LOFTIN R. B. & KENNEY P. J. Training the Hubble Space Telescope Flight
Team. IEEE Computer graphics and applications Vol 15, n° 5, September 1995 31-37 1995.
[LOU2003] Lourdeaux D., Mellet-d'Huart D. & Burkhardt J.-M. Potentialities of virtual
reality for pedagogical assistance Virtual concept 03, 2003.
[MAN2001] Mantovani F. VR learning: Potential and Challenge for the Use of 3D
Environments in Education and Training. In Riva G., Galimberti Eds. Toward
CyberPsychology: Mind Cognition and Society in the Internet Age. IOS Press, 2001.
[MEL2004] Mellet-d'Huart D. De l'intention à l'attention. Contributions à une démarche de
conception d'environnements virtuels pour apprendre à partir d'un modèle de l'(én)action. PhD
thesis from Université du Maine. 2004 – in French.
[MOS2002] Moshell M., Hugues C. & Loftin B. Virtual reality as a tool for academic
learning In STANNEY K. Ed. Handbook of Virtual Environments. Lawrence Erlbaum
Associates, 2002.
[NOE2000] Noe A. & O'rean K. Perception attention and the grand illusion. In Attentional
Blindness. Psyche 6(15) October 2000,
http://psyche.cs.monash.edu.au/v6/psyche-6-15-noe.html, Arien Mack's & Irvin Rock's Book,
2000.
[PAS1999] PASTRE P. L'ingénierie didactique professionnelle In CARRE P. & CASPAR P.
Eds. traité des sciences et des techniques de la formation. pp. 403-416, 1999 – In French.
[RIC2000] Rickel J. & Johnson W. L. Task-oriented Collaboration with Embodied Agents in
Virtual Worlds In Embodied Conversational Agents. MIT Press, 2000.
[RIC2002] Rickel J. W., Marsella S., Gratch J., Hill R., Traum D. & Swartout W. Toward a
New Generation of Virtual Humans for Interactive Experiences IEEE Intelligent Systems
July/August 2002, 2002.
[RIV2003] Riva G. Applications of virtual environments in medicine. Methods Inf Med
5/2003.
[WIN2002a] Winn W. What can students learn in virtual environments that they cannot learn
in class? In First International Symposium Open Education. Faculty, Andalu University,
2002.
[WIN2002b] Winn W., Windschitl M., Fruland R. & Lee Y.-L. Features of Virtual
Environments that Contribute to Learners' Understanding of Earth Science. 2002
[WIN2003] Winn W. Beyond constructivism: A return to Science-based research and practice
in Educational Technology, 2003.
AUTHORS

Dr Laurent Da Dalto is a senior engineer of the Virtual Reality Department of the CS

company. His main areas of expertise include physical lighting simulation, real-time
rendering and VR training. He is also the product manager of CS WAVE and the responsible
for the industrialisation of various VR training tools.
He received his Ph.D. in computer graphics from the Toulouse III university, France. He is
the author of 8 international papers and contributed to the redaction of several books in his
domain. Prior to this current position, he was the animation manager of the R&D team of
Kalisto Entertainment, a French video game company. There, he was in charge of the
development of the software engine for real time animation of bodies and structures (“Dark
Earth”, “The Fifth Element”, “Nightmare Creatures” video games).

Dr Olivier Balet is the Technical Director of CS' Virtual Reality Department. His main areas
of expertise include virtual reality, virtual storytelling, 3D interaction, multi-modal computer-
human interfaces, and cooperative working. Prior to his current position, he has conducted
research on 3D interaction at the Toulouse III university where he received his diploma
(M.S.) and Ph.D. in computer science (VR specialisation). He is the author and has been the
Project Manager of several European funded projects (CAVALCADE, VISIONS, V-Man, V-
Planet, VISTA, etc.).
He has been an active expert to the European Commission for activities under FP5 KA3 and
KA4 (Virtual Reality, Simulation, Multimedia and Collaborative Working Action lines) and
FP6 (FET, Cultural Heritage, networked audiovisual systems). He is VR assistant professor at
both the Toulouse III University and the ENSICA high school, program committee member
or reviewer for international conferences (Eurographics, ICVS, VRIS, Minitrack, TIDSE,
etc.), and the author of more than 25 papers published in international journals and conference
proceedings.

M. Dominique Steib is since 1998 project manager in the R&D "Industrial fields"
department of AFPA, the largest French vocational training organization for adults. He
introduced New Techonologies of Information and Communication in pedagogical resources
for industrial Afpa training. He is Afpa's project manager of CS WAVE. He participated in
conception team of other project RV training applications. He is expert in the field of
educational applications.
His educational background is a Master Degree of Psychopatologist and a training in Training
Ingenierie. He has worked with AFPA since 1977 as a labor psychologist, educational
adviser, training centre co-manager. training ingenierie. Since 1990 in Regional Direction he
participated in training and enemployement of disability people during 9 years also in
European projects and in Regional Direction he managed several innovations Afpa projects in
organisation and pedagogical training.

Dr Daniel Mellet-d'Huart is project manager in the R&D department of AFPA, the largest
French vocational training organization for adults. He introduced and developed virtual reality
systems to support learning and training. He participated in the realisation of industrial
applications such as CS WAVE, or others that are under development to train trainers to
interviews, or to train industrial machine operators. He is expert in the field of VR-learning:
design of training resources and educational applications. He is also an associate researcher in
CERV (European Center for Virtual reality) in Brest.
His educational background is a PhD in VR design method for learning applications, masters
in education, electronic media and management. He has worked with AFPA since 1982 as a
labor psychologist, educational adviser, training centre co-manager, member of a large
European project coordination team, forecasting engineer, and project manager for virtual
reality. He is the author of international papers, national papers, research reports and
participated in international scientific missions (http://www.dm-dh.com). He also co-authored
the VR learning chapter in FUCHS P. & MOREAU G. (Eds.) Le traité de la réalité virtuelle.
Presse de l'école des Mines de Paris, 2003,