This content is not included in
your SAE MOBILUS subscription, or you are not logged in.
Real-Time Prediction of Pre-ignition and Super-Knock in Internal Combustion Engines
Journal Article
03-16-03-0021
ISSN: 1946-3936, e-ISSN: 1946-3944
Sector:
Topic:
Citation:
Manzoor, W., Rawashdeh, S., and Mohammadi, A., "Real-Time Prediction of Pre-ignition and Super-Knock in Internal Combustion Engines," SAE Int. J. Engines 16(3):363-375, 2023, https://doi.org/10.4271/03-16-03-0021.
Language:
English
Abstract:
Super-knock is a phenomenon triggered by pre-ignition and has limited the design
envelope of internal combustion engines (ICEs) in terms of power density. This
poses a huge challenge for the automotive industry where engine sizes have been
continuously decreasing due to the demand for weight savings and integration
with electrified powertrains. Such downsized engines typically require increased
boost pressure, availing conditions conducive to pre-ignition, which in turn may
trigger super-knock. Traditionally, this and other forms of knock have been
managed by way of a “detection and mitigation” approach in place of “perdition
and avoidance” due to an evolving understanding of corresponding combustion
dynamics, as well as the incapability of emerging real-time computational
methods to perform and actuate over the timescale required. In this study, a
data-driven algorithm is used to extract (and adapt) a globally linearized
system representation using eigen-time-series, isolating the dynamic modes of
the system to capture underlying effects leading to pre-ignition without the
need for physics-based modeling. This approach is a unique application of the
“Hankel Alternative View of Koopman” (HAVOK) analysis for chaotic systems and
can be executed on board an engine control module supplying a buffer of recent
to latest time-step data to predict an impending pre-ignition event. The
proposed design does not require any change to existing sensors and actuators in
the existing knock management system architecture, nor would it require any
significant increase in computational capacity in terms of the associated engine
control unit. A simulation was conducted with real super-knock data to nominally
test the applicability of the algorithm. From this training dataset, the
linearized dynamic system was able to predict pre-ignition approximately 2.27 s
prior to the event, which is adequate to take mitigating action. Further
validation runs covering low, medium, and high engine speeds within the envelope
of low-speed pre-ignition (LSPI) generated similar results.