Title :
Dynamic mapping of diesel engine through system identification
Author :
Karlsson, M. ; Ekholm, K. ; Strandh, P. ; Johansson, R. ; Tunestal, P.
Author_Institution :
Dept. of Autom. Control, Lund Univ., Lund, Sweden
fDate :
June 30 2010-July 2 2010
Abstract :
From a control design point of view, modern diesel engines are dynamic, nonlinear, MIMO systems. This paper presents a method to find low-complexity black-box dynamic models suitable for model predictive control (MPC) of NOx and soot emissions based on on-line emissions measurements. A four-input-five-output representation of the engine is considered, with fuel injection timing, fuel injection duration, exhaust gas recirculation (EGR) and variable geometry turbo (VGT) valve positions as inputs, and indicated mean effective pressure, combustion phasing, peak pressure derivative, NOx emissions, and soot emissions as outputs. Experimental data were collected on a six-cylinder heavy-duty engine at 30 operating points. The identification procedure starts by identifying local linear models at each operating point. To reduce the number of dynamic models necessary to describe the engine dynamics, Wiener models are introduced and a clustering algorithm is proposed. A resulting set of two to five dynamic models is shown to be able to predict all outputs at all operating points with good accuracy.
Keywords :
MIMO systems; diesel engines; identification; nitrogen compounds; nonlinear dynamical systems; predictive control; MIMO systems; NOx emissions; Wiener models; black box dynamic models; clustering algorithm; control design; diesel engine mapping; dynamic systems; emissions measurements; exhaust gas recirculation; linear models; model predictive control; nonlinear systems; system identification; valve positions; variable geometry turbo; Control design; Diesel engines; Fuels; Geometry; MIMO; Nonlinear dynamical systems; Predictive control; Predictive models; System identification; Timing;
Conference_Titel :
American Control Conference (ACC), 2010
Conference_Location :
Baltimore, MD
Print_ISBN :
978-1-4244-7426-4
DOI :
10.1109/ACC.2010.5531242