Stimulation Initiative for European Neural Applications
Esprit Project 9811
Qualification of Shock-Tuning for Automobiles.
Company Background
Comar International B.V. is a small mechatronics company located in the center
of the Netherlands and operating on the international, highly competetive,
automotive testequipment market. The company has only limited internal research
resources.
The problem
Among the many factors that influence the safety of driving on public roads
is the technical maintenance level of the automobile itself. A regular car
inspection has therefore become a law-enforced reality. At the inspection site,
a number of dedicated apparatus is installed, on which a standardardized probing
of the safety limits is performed. To a lesser degree this also involves the
safety margins or, in other words, the comfort of driving. The correct tuning
of shocks has a major impact on the driving characteristics of an automobile.
However, the shocks take effect in combination with suspension stifness and
tyre hardness and hence current testmachinery can not reliably distinguish
between them. As effect it is estimated that at least 30% of the automobiles
of the public roads still have inferior shocks, which potentially can be the
cause of accidents. Several attempts have been made in the past to isolate
the effect of shock-tuning but to no avail.
Neural network application
Measurement data have been taken from a range of vehicles under different maintenance
conditions on an existing apparatus. After establishing that the information
content suffices for neural processing, a neural feedforward network has been
trained by error backpropagation and validated in the InterAct environment
to approximate the shock characteristics. This prototype function is ensuing
mathematically optimized for robustness and stability. The final network is
currently introduced in a new market offering to be sold on an international
scale.
Benefit
Neural networks allow a real-time correlation of measurement data to prototype
with various signal processing algorithms. In this manner a cost-effective
weighting algorithm has been found that adapts easily to different car types
and operating conditions. The mixed approach with neural rapid prototyping
and classical mathematical optimization techniques has proven to be very effective
for modelling non-algorithmic problems.
Generalization
The transition in technology base from mechanical to mechatronical through
the introduction of microsystems brings more than a simple cost reduction.
It also brings a sensitive set of data channels carrying influences from a
large number of sources. Instead of filtering out the signal as desired by
design, the quest is for a design in which many signals can be viewed in combination.
In conventional technology, mechanical micromachining has forced diagnosis
on a one feature per machine basis. Advances in microelectronics allow for
combined sensory systems in which the careful construction is replaced by novel
signal processing and optimal adaptivity by in-product training.
Contact Person
Dr.Ir. J.A.G. Nijhuis, Rijksuniversiteit Groningen, Dept. of Computing Science,
P.O. Box 800, NL-9700 AV Groningen (The Netherlands)