Research Collection Doctoral Thesis Optimal control of diesel engines modelling, numerical methods, and applications Author(s): Asprion, Jonas Publication Date: 2013 Permanent Link: https://doi.org/10.3929/ethz-a-010050065 Rights / License: In Copyright - Non-Commercial Use Permitted This page was generated automatically upon download from the ETH Zurich Research Collection. For more information please consult the Terms of use. ETH Library Diss. ETH No. 21593 OPTIMAL CONTROL OF DIESEL ENGINES Modelling, Numerical Methods, and Applications Jonas Asprion 2013 Jonas Asprion OPTIMAL CONTROL OF DIESEL ENGINES DISS. ETH Nr. 21593 OPTIMAL CONTROL OF DIESEL ENGINES: MODELLING, NUMERICAL METHODS, AND APPLICATIONS ABHANDLUNG zur Erlangung des Titels DOKTOR DER WISSENSCHAFTEN der ETH ZÜRICH (Dr. sc. ETH Zürich) vorgelegt von JONAS ASPRION MSc ETH in Mechanical Engineering geboren am 17.04.1984 von Escholzmatt, LU Angenommen auf Antrag von Prof. Dr. Lino Guzzella, Referent Prof. Dr. Moritz Diehl, Korreferent 2013 Jonas Asprion [email protected] © 2013 ETH Zurich Institute for Dynamic Systems and Control Sonneggstrasse 3 CH-8092 Zurich Switzerland When you truly love something you’re doing, you can’t do nothing wrong. — Ryan Dungey, 2010 To Marieken, Milla and Lilith. Leonhard Euler, Methodus inveniendi lineas curvas maximi minimive proprietate gaudentes sive solutio problematis isoperimetrici latissimo sensu accepti. Lausannae & Genevae: apud Marcum-Michaelem Bous- quet, 1744, p. 245. [http://dx.doi.org/10.3931/e-rara-1490] Nothing in the world takes place without optimization, and there is no doubt that all aspects of the world that have a ratio- nal basis can be explained by optimization methods.? ? Free translation by Martin Grötschel in the introduction of Optimization Stories, Documenta Mathematica, 2012. The original text is reproduced more closely by the translation provided in J.D. Barrow and F.J. Tipler, The Anthropic Cosmological Principle, Clarendon Press, 1986, on page 150: “For since the fabric of the universe is most perfect and the work of a most wise Creator, nothing at all takes place in the universe in which some rule of maximum or minimum does not appear ... there is absolutely no doubt that every affect in the universe can be explained satisfactorily from final causes, by the aid of the method of maxima and minima, as it can be from the effective causes themselves.” Acknowledgements During the past four years, my life often felt like an optimal control problem. The resource “personal energy” as a state variable needed to be wisely allocated to optimise several contrasting objectives and to sometimes work around seemingly infeasible constraints. Concurrently, the energy had to be regenerated regularly. The problem to be solved thus always was a problem of optimal time management. In this context, which is nicely captured by Leonhard Euler in his “De Curvis Elasticis”, I took the role of the optimisation algorithm that constantly tried to solve the problem “PhD thesis versus the rest of my life”. Without the help and the support from numerous persons, this algorithm would have had neither the necessary computational power available nor would it have converged sufficiently fast. Above all, I thank my wife Marieken. In the presence of “infeasible constraints”, she always agreed to slacken her demands to restore fea- sibility. Furthermore, she always helped me to recharge my personal energy – which also holds for our two children. However, the most im- portant role my wife takes is to keep showing me that some aspects of life do not belong to any optimisation problem but have to be enjoyed just as they come. Oscar Chinellato, my co-supervisor at FPT Motorenforschung AG, initiated the project “Model-Based Engine Calibration”. Besides pro- xi Acknowledgements viding this fundamental contribution, he kept asking critical questions and often provided ideas on how to tackle certain problems. The discus- sions with him were always sources of inspiration and motivation, and they helped to keep my “Newton steps” headed in the right direction. Another important person at FPT is Theo Auckenthaler, who initially “hired” me to perform my industrial internship as well as to do the re- search for my master thesis with the company. It was his persistent and successful pursuit of the model-based approach that paved the way for the realisation of such an innovative project in industry. The Institute for Dynamic Systems and Control laid the groundwork for this thesis. The lectures held by Prof. Lino Guzzella and Dr. Christo- pher Onder were some of the best I ever attended, and they inspired me throughout my bachelor and master studies. I am indebted to Prof. Lino Guzzella for providing me with the opportunity to pursue my doc- toral studies on a challenging topic. During my doctorate, he focussed on forcing my steps to maintain a reasonable length, even if I could not see any progress at that time. In retrospect, I appreciate this “school of hard knocks” – it forced me to make progress in leaps, not only from a technical point of view, but also personally. I also acknowledge the help I received from Brigitte Rohrbach who copy-edited all my publications. Her knowledge of the English language greatly helped me to improve the quality of my manuscripts. My co-referee, Prof. Moritz Diehl, had a significant impact on the course of my doctoral studies. In the spring term of 2011 he gave a lecture on “Numerical Optimal Control” at the Institute for Automation in the Department of Electrical Engineering. Until that time I had investigated mainly the heuristic type of optimisation strategies and had only recently discovered the power of the direct methods. His lecture encouraged me to focus my research concerning optimal control on this approach, which turned out to be the perfectly right choice. I am xii Acknowledgements grateful to him for bringing me into the right “region of attraction”. His fascination for the topic was contagious, and I greatly profited from his profound knowledge. Last but no least, my colleagues helped me throughout the time at the institute, some by technical or personal discussions, some by an auspicious collaboration. Others managed to act as an example for not taking this part of life overly serious or some step-enforcing measures from the supervisors too personal. In particular, I thank Stephan Zentner for all the joy and the sorrows we shared during our four common years as doctoral students. J. A. xiii Abstract In response to the increasingly stringent emission regulations and a demand for ever lower fuel consumption, diesel engines have become complex systems. The exploitation of any leftover potential during tran- sient operation is crucial. However, even an experienced calibration engineer cannot conceive all the dynamic cross couplings between the many actuators. Therefore, a highly iterative procedure is required to obtain a single engine calibration, which in turn causes a high de- mand for test-bench time. Physics-based mathematical models and a dynamic optimisation are the tools to alleviate this dilemma. This thesis presents the methods required to implement such an approach. Optimisation-oriented models for the air path and for the in-cylinder processes of diesel engines are derived, and the numerical methods re- quired to solve the corresponding large-scale optimal control problems are presented. The models are shown to be applicable to engines of var- ious sizes and configurations, and a transient experimental validation highlights their accuracy. A self-contained framework for numerical optimal control is implemented. The convergence behaviour and the computational performance of this framework when applied to the problem at hand are analysed. Finally, the resulting optimal control- trajectories over long driving profiles are shown to provide enough information to allow conclusions to be drawn for causal control strate- xv Abstract gies. Ways of utilising this data are illustrated, which indicate that a fully automated dynamic calibration of the engine control unit is conceiv- able. An experimental validation of the overall approach demonstrates the applicability of these results. The measurement results show that the dynamic optimisation predicts the reduction of the fuel consump- tion and the cumulative pollutant emissions with a relative error of around 10% on highly transient driving cycles. xvi Zusammenfassung Um die stetig strenger werdenden Emissionsvorschriften zu erfüllen und gleichzeitig dem Wunsch nach noch geringerem Treibstoffver- brauch nachzukommen, muss der transiente Betrieb von Dieselmoto- ren optimiert werden. Um das verbleibende Potential auszuschöpfen, stellt ein aktueller Dieselmotor viele, sich jedoch gegenseitig beeinflus- sende Aktuatoren zur Verfügung. Selbst ein erfahrener Applikationsin- genieur benötigt viele Iterationen, um eine akzeptable Bedatung aller Kennfelder und Reglerstrukturen für dieses komplexe System zu erlan- gen. Da komplette dynamische Fahrzyklen betrachtet werden müssen, resultiert ein immenser Bedarf an Prüfstandszeit. Das Ziel dieser Arbeit ist die Erstellung eines Werkzeugs, das den Ingenieur insbesondere für die Applikation des transienten Betriebes unterstützt und somit die kostenintensive Prüfstandszeit reduziert. Auf physikalischen Grundge- setzen bauende, mathematische Modelle für alle Prozesse im Motor sind dazu genauso nötig wie eine numerische Methodik, um schnell und verlässlich dynamische Optimierungsprobleme zu lösen. Die Mo- delle werden