Alexander Schreiber

Identification of Nonlinear Static Processes with Local Polynomial Regression and Subset Selection

Abstract The presented method for nonlinear system identification is based on the LOLIMOT algorithm introduced by Nelles and Isermann [1996]. The LOLIMOT algorithm divides the input space by a tree-construction algorithm and interpolates the local linear models by local membership functions. Instead of assuming local linear models, the presented algorithm utilizes general local nonlinear functions, which make the algorithm more flexible. These are approximated by a multidimensional Taylor series. Since the amount of regressors grows fast with the number of inputs and the expansion order, a subset selection procedure is introduced. It reveals significant regressors and gives information about the local functional behavior. The local subset selection is implemented as a stepwise regression with replacement of regressors. Mallows’ C p -statistic is used for the subset selection algorithm and is also implemented for final model selection. The benefit of the extended algorithm lies in the higher flexibility in the local models, which results in less partitions of the input space by a similar approximation quality.

Identification methods for experimental nonlinear modeling of combustion engines

Abstract The optimized control of combustion engines with regard to minimized fuel consumption and emissions requires nonlinear models. Because of an increase of control inputs, like fuel mass flow, injection angle, exhaust gas recirculation flow and several outputs like torque, nitrogen oxides (NOx), hydrocarbons (HC) and particulates the classical grid-based measurement techniques take too long time and do not include dynamics. Therefore different measurement strategies for the stationary and dynamic behavior are described, like Design of Experiments (DoE) and use of suitable neural networks and Pseudo-Random-Binary-Signals (PRBS). As the structure of the models is not precisely known a-priori, nonlinear identification methods in form of special versions of neural networks are good candidates. Therefore, it will be shown how with special amplitude-modulated pseudo random binary signals (APRBS), simultaneous excitation of several input signals, nonlinear multi-input multi-output models can be obtained in relatively short time.

Dynamische Motorvermessung mit verschiedenen Methoden und Modellen

Die stark zunehmenden gesetzlichen und wirtschaftlichen Vorgaben zur Senkung von Kraftstoffverbrauch und Abgasemissionen stellen grose Anforderungen an die weitere Entwicklung von Benzin- und Dieselmotoren. Hierbei sind grundlegende Fortschritte durch Konstruktion und auslegungsbedingte Masnahmen im Bereich der Einspritzung, Gemischaufbereitung, Aufladung, Brennverfahren und Abgasnachbehandlung zu erreichen. Ein wesentlicher Teil dieser Verbesserungen wird jedoch durch eine Zunahme von Variabilitaten erreicht wie z.B. verstellbaren Vor-, Haupt- und Nacheinspritzungen, variablem Raildruck, variablen Nockenwellensteuerwinkeln, Ventilhuben, Drall-/Tumbleklappen sowie verstellbaren Abgasturbinen, Abgasruckfuhrstromen und Abgasnachbehandlungssystemen. Dadurch steigt die Zahl der Stellglieder (Aktoren) stark an. Hinzu kommen zusatzliche Sensoren wie z.B. fur Luftzahl, NOx, Brennraumdruck, Abgastemperatur und Abgasdruck. Deshalb nimmt der Umfang der Steuerungs-, Regelungs- und Diagnosefunktionen in der Motorelektronik (ECU) stark zu. Bild 7-1 zeigt als Beispiel den Signalfluss fur die gesteuerten und geregelten Grosen eines Dieselmotors in einer beispielhaften Prufstandsumgebung.

Identification of the nonlinear, multivariable behavior of mechatronic combustion engines

Abstract The identification of the nonlinear dynamic behavior of combustion engines on test benches is described. A method, that is based on the combination of statistical stationary engine measurement (Design of Experiments - DoE) and dynamic identification techniques for nonlinear multi-input systems is shown. The dynamics are excited by several amplitude-modulated pseudorandom-binary-signals (APRBS), which act on the engine simultaneously, in order to allow a short measurement time period. To show the applicability of the described identification and modelling methods several models for a common-rail diesel engine with exhaust gas recirculation and variable geometry turbocharger (VGT) are shown, with five input variables and the outputs nitrogen-oxide emissions, opacity, air-fuel ratio (Lambda) and exhaust gas temperature.

Identification of Combustion Engines With Special Orthogonal Excitation Signals

This article shows a new excitation method for the identification of combustion engines. The method is based on the combination of statistical stationary engine measurement (Design of Experiments - DoE) and dynamic identification techniques for nonlinear multi-input systems. The dynamics are excited by special amplitude-modulated pseudo-random-multilevel-signals (APRMS), which act on the engine simultaneously, in order to allow a short measurement time period. Because the used dynamic excitation signals are orthogonal, they allow a better separation of the input signals for system identification.© 2010 ASME