IAV Automotive Engineering and Calico Systems Announce the Formation Of Automotive Neural Network Consortium (AutoNNC)

    ANN ARBOR, Mich., April 26 IAV Automotive Engineering and
Calico Systems announce the formation of an automotive research and
development consortium aimed at the widespread deployment of neural network
software technologies in powertrain applications.  The Automotive Neural
Network Consortium (AutoNNC) will conduct a four-year program of pre-
competitive research in the emerging area of the application of artificial
neural networks (ANNs) to the control, calibration, diagnosis, simulation and
modeling of advanced powertrain systems.  Targeted applications will include
light-duty gasoline and diesel vehicles, medium and heavy-duty diesel engines,
hybrid electric vehicles and other emerging technologies such as fuel cells.
    Given sufficient data, neural networks (which are believed to mimic the
manner in which the human brain operates) have the capability to learn the
complex, multidimensional and non-linear relationships between many
independent and dependent variables.  For example, given sufficient data on
real engine performance during training, a neural network model can be
developed that shows the relationships between engine performance, individual
exhaust gas emissions rates and engine operating parameters in real-time,
across any transient operating cycle.  Once developed, this model can then be
used for real-time optimal engine control, for on-line diagnostics, for off-
line engine calibration or for engine simulation.
    Ever-stricter emission regulations and increasing demands for improved
vehicle fuel economy have resulted in tremendous complexity in automotive
powertrain control.  With the explosive proliferation of emerging engine and
exhaust aftertreatment technologies, there is apparently no end in sight to
this trend.  The targeted use of neural networks in powertrain applications
has the potential to reduce software complexity enormously while adding
significant new functionality.  "The replacement of a physical model with a
neural network-based algorithm for charge air determination in a new high
complexity gasoline engine application resulted in a significant reduction in
the time required to code and calibrate a new engine control strategy," said
Sven Meyer, Senior Control Systems Engineer at IAV Inc.
    The potential high value applications of ANNs to powertrain development
are not just in the area of engine control.  According to Dr. Chris Atkinson,
Chief Engineer of Calico Systems, "Recent successes in expediting engine and
vehicle calibration using neural network modeling-based techniques have opened
up a whole new avenue of approach for reducing vehicle time-to-market.  Neural
network-based rapid calibration methods have reduced the time required to
calibrate a five parameter diesel engine control system with over 2700
discrete calibrateable parameters from 12 months to 3 months."
    The Consortium is targeted primarily at automotive manufacturers, engine
manufacturers, powertrain component and system suppliers, and engine control
and calibration tool suppliers.  The Consortium will conduct projects of
relevance to both gasoline and diesel powertrain engineers by employing a dual
track approach.  All work conducted will be non-vehicle or engine platform
specific and, as an alternative, will use state-of-the-art generic engine
control hardware and engines.  The technology targeted will be 2005-2007
emissions standards compliant.
    The pre-competitive research focus of the Consortium will be to
investigate the use of neural network-based techniques to

    *  Reduce powertrain software complexity, while accommodating new
technologies and approaches
    *  Reduce time-to-market for new powertrain control systems
    *  Reduce significantly the time, effort and costs required to calibrate
engines and vehicles for emissions, fuel consumption and driveability
constraints
    *  Develop new on board diagnostic techniques
    *  Reduce hardware requirements with virtual sensing technologies
    *  Develop fully optimized, transient neural network model-based mapless
engine control systems

    The benefits to participation in the Consortium will include the
identification of potential smart applications of ANNs in next generation
powertrain activities, the development of industry standards for the
integration of ANN techniques in powertrain control and calibration, and
access to reliable results from case studies and real-world investigations of
successful ANN implementation.  It is widely anticipated that this decade will
see the widespread deployment of ANNs in select powertrain control,
calibration, diagnostics, modeling and simulation applications, and this
Consortium aims to facilitate that goal.
    The Automotive Neural Network Consortium (AutoNNC), which will commence in
July 2001, will be co-hosted by IAV Automotive Engineering Inc., and Calico
Systems Inc.
    For more information on the Automotive Neural Network Consortium, go to
http://www.autonnc.org

                    MAKE YOUR OPINION COUNT -  Click Here
               http://tbutton.prnewswire.com/prn/11690X07364158