This content is not included in
your SAE MOBILUS subscription, or you are not logged in.
Improved Diesel Engine Load Control for Heavy-Duty Transient Testing Using Gain Scheduling and Feed-forward Algorithms
Journal Article
03-16-06-0042
ISSN: 1946-3936, e-ISSN: 1946-3944
Sector:
Topic:
Citation:
Cook, J., Puzinauskas, P., Wagenmaker, M., and Bittle, J., "Improved Diesel Engine Load Control for Heavy-Duty Transient Testing Using Gain Scheduling and Feed-forward Algorithms," SAE Int. J. Engines 16(6):739-761, 2023, https://doi.org/10.4271/03-16-06-0042.
Language:
English
Abstract:
Heavy-duty (HD) engines for sale in the United States must be demonstrated to
emit below allowable criteria and particulate emission limits over the
operational load and speed cycle specified by the Federal Test Procedure (FTP)
Heavy-Duty certification test. The inherently nonlinear load response of
internal combustion engines tends to increase torque variability during the most
dynamic portions of the test cycle. This clouds assessment of engine
developments intended to improve transient performance and leads to frequent
invalidation of certification tests. This work sought to develop and evaluate
test torque control strategies that reduce this variability. Several
load-control algorithms were evaluated for this purpose using a Cummins ISX15 HD
diesel engine loaded with a transient alternating current (AC) dynamometer. The
evaluations were performed over an abbreviated version of the FTP test which
included the most transient portions of the drive cycle and were based on the
regression characteristics specified by the EPA. These algorithms included a
basic proportional-integral (PI) feedback algorithm, which was designated as the
baseline for this work, an open-loop feed-forward-only (OLFF) algorithm, a
gain-scheduled PI (GSPI) algorithm, a combined feed-forward and PI (FFPI)
algorithm, and a gain-scheduled feed-forward PI (GSFF) algorithm. Each algorithm
uniquely changed the torque control depending on the specific part of the cycle
observed and the metric used to quantify performance; however, the FFPI and the
GSFF had significantly better overall performance than the other algorithms
tested. The FFPI had the greatest performance in regression intercept,
coefficient of determination (R2), and standard error of estimate
(SEE), with 70.1%, 7.3%, and 56.0% respective improvements relative to the
baseline algorithm. The GSFF algorithm demonstrated the greatest regression
slope performance with an improvement relative to the baseline of 23.0%. Such
improvement not only reduced the number of tests invalidated by exceeding
regression limits, but the associated improvement in repeatability could allow
more accurately quantifying engine and aftertreatment development impacts on
transient performance. Such improved development efforts should help facilitate
closing the gap between transient and steady-state performance, leading to
efficiency improvements and greenhouse gas and criteria emission reductions.