Conference article

This paper describes splitting a crank angle resolved three cylinder combustion engine with an air path model and a combustion model. This is to distribute the computational effort on hardware by running models on separate cores to achieve real time capability. Hardware tests show the split models are not able to achieve real time because the thermal dynamics of air path model and combustion model are highly interconnected and computing the models on separate cores will introduce delay and solution can become inaccurate and even infeasible. In order to achieve real time capability while ensuring the results are accurate (2-5% percent max. error), a new method is proposed, in which instead of running with a complete fluid intake and exhaustmodel, the combustion model runs with a mean value intake model calibrated for many operating points across the speed-load range. The results show that the combustion model running with mean value intake model is able to produce highly accurate result and real time capability is achievable. By using mean value intake model, calibration effort is significantly reduced compared to purely table based method as the mean value model captures essential dynamics and is able to predict reliably between transition from one operating point to another. The mean value method takes into account Air Fuel Ratio (AFR) dynamics and thus calibration against AFR becomes unnecessary. Comparing to a non-mean value purely table based method, the latter requires calibration at densely scattered operating points in order for the transition between each calibration point to be smooth enough. In calibrating the mean value model a controller is designed to control the dynamics error to zero. This control based method shows high efficiency compared to optimization tools as it does not depend on initial values and iteration process of the calibrating parameters. A function is created to automatically create the tables calibrated. The calibrated mean value intake model is run with a combustion model on a Concurrent test/HiL rig and shows real time capability is achieved with good accuracy. The physical engine model is built in Dymola.

mean value intake model, split engine model, automated calibration

L. Guzzella and C.H. Onder. Introduction to modelling and control of internal combustion engine systems. Springer-Verlag Berlin Heidelberg, 2004. ISBN 978-3-642-10775-7.

X. Han. Splitting mechanical, fluid ports and thermal ports usingreal inputs and real outputs. http://www.claytex.com/techblog/splitting-mechanical-and-fluid-devices-using-realinputs-real-outputs/, 2017a.

X. Han. Linearised Mean Value Engine Model-based Idle Control.
http://www.claytex.com/blog/linearised-mean-valueengine-model-based-idle-control/, 2017b.

X. Han. Idle speed control using model based sliding mode control. http://www.claytex.com/blog/idle-speed-control-usingmodel-based-sliding-mode-control/, 2017c.

J.B. Heywood. Internal combustion engine fundamentals. McGraw-Hill, Inc, 1988.