GT-POWER Real-Time – Diesel engine model for

Hardware in the Loop testing

GT-SUITE Conference 2009

David Willermark, SEMCON

Niklas Smith, Volvo Car Corporation

www.volvocars.com

Contents

�Project background

�Introduction to Mean Value/RT modeling

�Model details and results

�Introduction to Hardware in the Loop (HiL)

�HiL implementation

�Summary

�Project background

�Introduction to Mean Value/RT modeling

�Model details and results

�Introduction to Hardware in the Loop (HiL)

�HiL implementation

�Summary

Project background

�Volvo Cars initiated a project to evaluate a GT-POWER Real-Time

engine model for Hardware in the Loop application

�Semcon Gothenburg was involved to develop the engine model

�The new Volvo 5-cyl, two stage turbo, Euro5 diesel engine was

chosen for this study

�Volvo Cars initiated a project to evaluate a GT-POWER Real-Time

engine model for Hardware in the Loop application

�Semcon Gothenburg was involved to develop the engine model

�The new Volvo 5-cyl, two stage turbo, Euro5 diesel engine was

chosen for this study

www.volvocars.com

�The main purpose of a Mean Value model is to increase simulation

speed

�Derived from a detailed GT-POWER model

�Provides cycle averaged data

�Semi predictive

�Neural networks - trained by data either from the detailed model

or engine test

�The RT model is created from the MeanValue model and should

give identical results

�The main purpose of a Mean Value model is to increase simulation

speed

�Derived from a detailed GT-POWER model

�Provides cycle averaged data

�Semi predictive

�Neural networks - trained by data either from the detailed model

or engine test

�The RT model is created from the MeanValue model and should

give identical results

Mean Value/RT modelling – general description

Model conversion

Detailed model Mean value model

Inlet manifold

Cylinders

Exhaust manifold

�Flow components are lumped together to larger

volumes

�The cylinders are exchanged with a MeanValue

cylinder which is controlled by neural networks

�Flow components are lumped together to larger

volumes

�The cylinders are exchanged with a MeanValue

cylinder which is controlled by neural networks

MeanValue model details

�Neural networks - trained with data from the detailed model

- Volumetric efficiency

- IMEP

- Exhaust temperature

- FMEP

- Heat exchanger temperatures

�Detailed model DOE - number of points chosen according to the

guidelines from Gamma

� Turbine efficiency adjustments necessary

�Neural networks - trained with data from the detailed model

- Volumetric efficiency

- IMEP

- Exhaust temperature

- FMEP

- Heat exchanger temperatures

�Detailed model DOE - number of points chosen according to the

guidelines from Gamma

� Turbine efficiency adjustments necessary

Model results - Steady stateDeviation MeanValue vs. detailed model in %

Inlet manifold pressure Exhaust manifold pressure

Most points are within 2% - acceptable agreement

Model results - Steady stateDeviation MeanValue vs. detailed model in %

Air massflow Exhaust manifold temperature

Most points are within 2% - acceptable agreement

Transient results, NEDC

Introduction to HiL

�Engine control systems are getting more and more complex

and a larger number of variants is required

- Need for early development of engine control software

and hardware

�Hardware in the Loop (HiL) simulations

-Testing of control system related hardware, for example

the ECU, together with simulation models of its

surrounding systems (engine, actuators, sensors, vehicle

etc.)

- Development and testing of system components before

prototypes of the surroundings exist

�Engine control systems are getting more and more complex

and a larger number of variants is required

- Need for early development of engine control software

and hardware

�Hardware in the Loop (HiL) simulations

-Testing of control system related hardware, for example

the ECU, together with simulation models of its

surrounding systems (engine, actuators, sensors, vehicle

etc.)

- Development and testing of system components before

prototypes of the surroundings exist

HiL system overview

GT-RT model�Runs on separate board

HiL rig

www.dspace.com

Simulink model�Transmission

�Controls

ECU

HiL – implementation

�The HiL tests were run on a dSpace rig with a DS4505 board for the

RT model

�Smooth and troublefree installation of the board

�No application related operating problems

�Most effort was spent on arranging the interface between engine

model and vehicle model

�Testrun with real ECU and vehicle model was performed with

satisfying results

�The HiL tests were run on a dSpace rig with a DS4505 board for the

RT model

�Smooth and troublefree installation of the board

�No application related operating problems

�Most effort was spent on arranging the interface between engine

model and vehicle model

�Testrun with real ECU and vehicle model was performed with

satisfying results

Summary and conclusions

�The MeanValue/RT model results have acceptable agreement with

the detailed model in both steady state and in transient operation

�GT-power RealTime is a simple method to create good working

engine models for HiL applications

�Considering the limited work effort, the overall results are

satisfying

�The MeanValue/RT model results have acceptable agreement with

the detailed model in both steady state and in transient operation

�GT-power RealTime is a simple method to create good working

engine models for HiL applications

�Considering the limited work effort, the overall results are

satisfying

Acknowledgements

�Iakovos Papadimitriou and Matt Warner at Gamma Technologies

�Sverker Hansson, Fredrik Ragnarsson, Jan Styffe and Peter Norin at

Volvo Car Corporation

�Tino Schulze at dSpace

�Iakovos Papadimitriou and Matt Warner at Gamma Technologies

�Sverker Hansson, Fredrik Ragnarsson, Jan Styffe and Peter Norin at

Volvo Car Corporation

�Tino Schulze at dSpace

Contact: david.willermark@semcon.com