C.E.A. Finney, K. Nguyen
University of Tennessee
Knoxville TN 37996-2210
We present a method based on a maximum-likelihood estimator of Kolmogorov entropy to quantify the degree of cycle-to-cycle variation in internal combustion engines. This estimation can be applied to evenly sampled, continuous time series, such as cyclinder pressure traces, from which time-embedded trajectories can be reconstructed. Because the estimator is applied to reconstructed trajectories, it can account for nonlinear structure and high dimensionality in the measured variable.
We suggest that cycle-to-cycle combustion differences can be highlighted by focusing on windows that define trajectory segments representing significant combustion events and ignoring the remaining portions of the cycle. This windowing scheme removes motoring effects or bias during the exhaust and intake strokes in which certain thermodynamic variables are constrained by the physical motion of the piston in the cylinder and not by combustion effects.
We demonstrate the applicability of this windowed entropy estimator to experimental time series of pressure measurements of a four-stroke, one-cyclinder industrial spark-ignition engine.
Please email comments about this page to email@example.com.
This page's URL is http://www-chaos.engr.utk.edu/abs/abs-jcs32.html.
Updated: 2001-05-15 ceaf