OPEN FORUM
Safety Certificate: an audification performance of high-speedtrains
Florian Grond
Received: 13 November 2010 / Accepted: 9 August 2011 / Published online: 26 August 2011
� Springer-Verlag London Limited 2011
Abstract Safety Certificate is a musical performance
based on sensor data from high-speed trains. The original
purpose of this data is to provide a basis for the assess-
ments of the mechanical aspects of train safety. In this
performance, the data, which represents dynamical pro-
cesses below the audible range, are converted into sound
through audification. The sound that is generated live
during the performance is manipulated through the Manta
control interface, which allows for the convenient layering
of 48 different timbres. Safety Certificate was premiered at
Seconde Nature in Aix-en-Provence in March 2010 during
the Sonification symposium–What, Where, How, Why,
organized by Locus Sonus. The following short article
gives details about the data, the audification technique, use
of the control interface, and the musical structure of the
performance.
Keywords Audification � Performance
1 Introduction
Traveling with trains is an experience that has always had a
strong acoustic dimension. Sonic associations of trains can
be as old fashioned as the steam blow from a nineteenth
century train, or more subtle like the characteristic repeti-
tive sound of the wheel hitting the gap between two joint
rails. Today, the comfort of traveling in the acoustically
well-insulated wagons of high-speed trains has deprived us
of many of those sonic experiences.
In the sonification discourse, it has often been argued
that listening is a powerful way of assessing the proper
functioning of mechanical devices. In practice, technicians
often use their listening skills as a diagnostic device before
disassembling an engine. This inference based on acoustic
cues is also similar to medical diagnosis that is done by
listening to a patient’s chest.
High-speed trains need to undergo intensive testing
during the homologation process, before they are approved
for public service. This process entails the collection of
data from measurements of the mechanical parts of the
train. Interestingly, it is not only the train passengers who
hear little of the vehicle transporting them; technicians who
measure and maintain the trains also rely almost exclu-
sively on vision. Assessment of the wear of mechanical
parts of the train, such as wheels and axles, is an involved
process, and the interpretation of results relies on the visual
representation of charts and numbers. Listening to mea-
surement data, however, can reveal a great deal about the
forces that act onto the mechanical parts and their dynamic
response. In the data used in this performance, the slow
rocking movement of a wagon going at constant speed or
the harsh force of the brakes, when the train decelerates,
can be clearly heard. This suggests that an auditory display
of these data might have benefits for monitoring purposes.
2 The data
Data used in this performance are measurements from
sensors, which are attached to mechanical parts of high-
speed trains. These data are usually used as the basis of
subsequent sophisticated simulations, which allow
F. Grond (&)
CITEC Cognitive Interaction Technology
Centre of Excellence, Bielefeld University,
Universtatsstrasse 21-23, 33615 Bielefed, Germany
e-mail: [email protected]
123
AI & Soc (2012) 27:293–295
DOI 10.1007/s00146-011-0351-5
conclusions to be drawn about the durability of important
parts critical for the safety of the train. For Safety Certifi-
cate, three different data channels were used.
The first data channel corresponds to the sensor mea-
suring the vertical acceleration of the wheel. This sensor is
directly attached to the axle box. The second channel is the
measurement of the vertical displacement of the wheel. This
up and down movement corresponds to the extension of the
shocks of the wheel suspension. The third data channel is
the rotation around the vertical axis of the wagon. This can
be though of as the movement of a reversed pendulum
pointing from the rails up. The first data channel is partic-
ularly interesting, since the axle box has no suspension and
hence has to directly absorb all forces acting on the wheel.
While sitting in the wagon, a passenger is removed from
this direct impact by two layers of suspension.
The data are all originally recorded at 1,200 Hz and are
low-pass filtered at 300 Hz, because for the homology
process and dynamic processes in this frequency range are
of interest. This low-frequency range is due to the fact that
that the moving parts of trains have a heavy mass, and
therefore can only oscillate within low frequencies. The
highest perceived frequencies of dynamical phenomena
captured by these data appeared to be oscillations of 60 Hz.
The average frequency is, however, much lower and the
biggest part of the data, therefore, lie in a very subfrequent,
almost inaudible range.
3 Audification
The subfrequency data from real physical processes as
described above are an ideal candidate for audification.
Among all sonification techniques, audification is the most
direct conversion of data into sound.
More specifically, the data is loaded into a buffer and the
only further data transformation is amplitude normalization.
Data in the buffer is then played directly by sending it to the
digital to audio converter. The most important degree of
freedom for manipulating the sound in audification is the
change of playback speed. This is in fact necessary since the
processes measured from the trains are not happening on a
time scale that would lead to sound signals within the
human audible range. Data from the sensor that was directly
attached to the axle box was the primary source for audi-
fication, since it contained the most sonically interesting
features. The other two channels were used to control the
spatialization of the sound. For the purpose of spatialization
in the stereo field, the playback speed differed from that
used for audification. Hence, the movement of the sound
was not synchronous with the audification itself. This
artistic intervention was necessary in order for the spatial-
ization of the sound to take place at perceivable speed.
4 The control interface
For the real-time control of the sound during the perfor-
mance, I used the Manta,1 a touch-sensitive interface for
controlling music or video. The Manta consists of 48
sensors in a hexagonal array. These sensors measure
the resistance of the skin, which can be modulated by the
surface area your finger is covering. The sensors of the
touch-sensitive interface also features LED backlighting,
which, when activated, provides the performer with visual
feedback. For the audience, this backlighting evoked the
illusion of a real desk in a control center, where many
trains are simultaneously monitored. A photo of the inter-
face as used during the performance is depicted in Fig. 1.
For sound synthesis, the Manta was connected to Super-
Collider McCartney (1996). Each of the 48 sensors on the
interface controlled the volume of one audification–all 48
playing simultaneously. Each audification is tuned to a
different playback speed. The playback speed was loga-
rithmically mapped between the original data rate to a 5
times that rate, taking into account the non-linear relation
of frequency to pitch. In addition to the panning defined by
data, the amplitude of playback was modulated via sensor
values of the Manta. Low pressure leads to less volume and
more spatial movement, whereas strong pressure meant
louder playback with almost no stereo panning.
5 The performance
Safety Certificate stands in the long tradition of using train-
related sounds for music, for instance, the Mouvement
Symphonique Pacific 231 by Arthur Honegger or the Etude
aux Chemins de Fer by Pierre Schaeffer. Inspired by these
classic works, the performance follows the structure of a
Fig. 1 The control interface Manta used by the author during the
concert in Aix-en-Provence
1 The Manta is build by Jeff Snyder see http://www.snyder
phonics.com/.
294 AI & Soc (2012) 27:293–295
123
sonata movement. I usually inform the audience that they
will be listening to data, without specifying their origin.
The audience know that they hear first the data at two
times the original speed, in classical terms, this corre-
sponds to the exposition of the theme. I chose not to use the
original playback speed in the exposition, because this very
slow playback causes some very deep and often-unnoticed
rumbles in the subfrequency range, an experience that can
be better appreciated after familiarization, through careful
listening, to the audification of the data. This is why I put
the original playback speed at the end of the performance,
when the theme will be recapitulated.
The development of the theme starts at low frequencies
and successively layers different playback speeds resulting
in an ever-changing configuration of timbres. As mentioned
above, slow playback speeds tend to produce subaudio
frequencies, and it is only if one knows the origin of the data
that any similarity with the almost inaudible rocking
movement of trains can be perceived. At medium playback
speed, the sound is more reminiscent of a train. When it is
speeded up even further, the data sounds like a ski lift and
then like a tiny clockwork mechanism. The surprising
variety of different timbres that arise from the audification
of these data sets is the material for the musical counter-
point during the developments of the theme.
6 Conclusion
Safety Certificate is a life performance that combines
technological information from high-speed trains with
classical musical forms. Live audification together with an
esthetically appealing control interface proved to be a
successful combination in terms of creating an evocative
experience through the performance. Beyond this, the
surprising richness of the sounds produced suggest that
audification of the data might be useful for the monitoring
of mechanical train parts.
Acknowledgments The author would like to thank Fabian Schmid
from PJ Messtechnik (http://www.pjm.co.at) for providing the data of
the high-speed trains. Till Bovermann gave me a very useful intro-
duction howto program the Manta interface, which is a SuperCollider
class implemented by Alberto deCampo.
Reference
McCartney J (1996) Supercollider: a new real-time synthesis
language. In: Proceedings of international computer music con-
ference (ICMC’96), pp 257–258
AI & Soc (2012) 27:293–295 295
123