Turbocharger Modeling for Automotive Control Applications
Ford Forschungszentrum Aachen
Ford Research Laboratories
Copyright © 1999 Society of Automotive Engineers, Inc.
Dynamic simulation models of turbocharged Diesel andgasoline engines are increasingly being used for designand initial testing of engine control strategies. The turbo-charger submodel is a critical part of the overall model,but its control-oriented modeling has received limitedattention thus far. Turbocharger performance maps aretypically supplied in table form, however, for inclusion intoengine simulation models this form is not well suited.Standard table interpolation routines are not continuouslydifferentiable, extrapolation is unreliable and the tablerepresentation is not compact. This paper presents anoverview of curve fitting methods for compressor and tur-bine characteristics overcoming these problems. Weinclude some background on compressor and turbinemodeling, limitations to experimental mapping of turbo-chargers, as well as the implications of the compressormodel choice on the overall engine model stiffness andsimulation times.The emphasis in this paper is on compressor flow ratemodeling, since this is both a very challenging problemas well as a crucial part of the overall engine model. Forthe compressor, four different methods, including neuralnetworks, are presented and tested on three differentcompressors in terms of curve fitting accuracy, modelcomplexity, genericity and extrapolation capabilities.Curve fitting methods for turbine characteristics are pre-sented for both a wastegated and a variable geometryturbine.
INTRODUCTION AND BACKGROUND
Dynamic simulation models of turbocharged Diesel andgasoline engines are increasingly being used for designand initial testing of engine control strategies. The turbo-charger submodel is a critical part of the overall model,but its control-oriented modeling has received limitedattention thus far. A standard approach still appears to beto include the turbocharger performance data in the formof lookup tables directly into the model , . However,this form is not ideally suited for use in control-orientedengine models because the standard linear interpolationroutine is not continuously differentiable, sometimes lead-ing to apparent discontinuities in simulations. Further-more, and more seriously, this type of model does notadequately handle operating conditions outside of themapped data range, for example at very low turbochargerrotational speeds.While engine mapping usually covers the entire operatingrange, the situation for the turbocharger unit is different.Generally, it is possible to obtain the performance charac-teristics from the supplier. However, the turbochargercharacteristics are typically only mapped for higher turbospeeds (typically 90000 RPM and up) and pressureratios, whereas the operating range on the engine ex-Figure 1.Typical compressor map. Usually, such acompressor map shows constant speedlinesand constant isentropic efficiency lines. Here,we have omitted the efficiency lines andinstead superimposed the compressor flowrate determined from engine mapping data. Itis readily apparent that the compressormapping data does not cover the operatingrange of the compressor on the engine. Inparticular, low speed and low pressure ratiodata are lacking.
00.050.10.150.18.104.22.16822.214.171.124.8126.96.36.199.890000120000140000160000180000Scaled compressor flow parameter
p r e s s u r e r a t i o
Compressor mapMeasured engine data