Search Results

Advanced engine control and diagnosis strategies for internal combustion engines need accurate feedback information from the combustion engine. The feedback information can be utilized to control combustion features which allow the improvement of engine's efficiency through real-time control and diagnosis of the combustion process. This article describes a new method for combustion phase and IMEP estimation using one in-cylinder pressure and engine speed. In order to take torsional deflections of the crankshaft into account a gray-box model of the crankshaft is identified by subspace identification. The modeling accuracy is compared to a stiff physical crankshaft model. For combustion feature estimation, the identified MISO (multiple input single output) system is inverted. Experiments for a four-cylinder spark-ignition engine show the superior performance of the new method for combustion feature estimation compared to a stiff model approach.

Nowadays automobiles are getting smart and there is a growing need for the physical behavior to become part of its software. This behavior can be described in a compact form by differential equations obtained from modeling and simulation tools. In the offline simulation domain the Functional Mockup Interface (FMI) [3], a popular standard today supported by many tools, allows to integrate a model with solver (Co-Simulation FMU) into another simulation environment. These models cannot be directly integrated into embedded automotive software due to special restrictions with respect to hard real-time constraints and MISRA compliance. Another architectural restriction is organizing software components according to the AUTOSAR standard which is typically not supported by the physical modeling tools. On the other hand AUTOSAR generating tools do not have the required advanced symbolic and numerical features to process differential equations.

Within the scope of this work, the convective mass transfer to the zirconia sensor element of an exhaust oxygen sensor was analyzed experimentally and numerically. For the experimental setup, a heightened model of an oxygen sensor was built from Lucite® considering the similarity theory. Mass transfer is measured based on the absorption of ammonia and subsequent immediate color reaction. For the numerical investigation, a three-dimensional model of the test rig was built. To predict the flow pattern and the species transport inside the protection tubes, the commercial CFD-Code FLUENT® is used. The model for the mass transfer to the surface is implemented through user-defined functions.

In order to reduce the spark-ignition engine exhaust-gas emission and fuel consumption, it is essential that the required air/fuel ratio is maintained under all operating conditions. An important contribution to this claim is delivered by the injection valve by metering the fuel precisely and producing fine atomization. In this report experimental methods to get specific measuring information and methods for optimizing flow in injection valves are described. Original valves as well as large-scale models were used for the investigations concerning the steady and unsteady-flow characteristics, and were equipped with a number of different sensors. Holograms of the short-time recording of the spray cone are generated and used for the quantification of the atomization quality when injecting into atmospheric pressure and into vacuum, thus complying with the conditions encountered in the engine intake-manifold.

Compressed Natural Gas (CNG) is a promising alternative fuel for internal combustion engines as its combustion is fuel-efficient and lean in carbon dioxide compared to gasoline. The high octane number of methane gives rise to significant increase of the thermodynamic efficiency due to higher possible compression ratios. In order to use this potential, new stratified mixture formation concepts for CNG are investigated by means of numerical fluid simulations. For decades RANS methods have been the industry standard to model three-dimensional flows. Indeed, there are well-known deficiencies of the widely used eddy viscosity turbulence models based on the applied Boussinesq hypothesis. Reynolds stress turbulence models as well as scale resolving simulation approaches can be appealing alternative choices since they offer higher accuracy. However, due to their large computing effort, they are still mostly impractical for the daily use in industrial product development processes.

Modern zirconia oxygen sensors are heated internally to achieve an optimal detection of the oxygen concentration in the exhaust gas and fast light off time. The temperature of the gas in the exhaust pipe varies in a wide range. The zirconia sensor is cooled by radiation and forced convection caused by cold exhaust gas. If the zirconia temperature falls, the oxygen detection capability of the sensor decreases. To minimize the cooling effects, protection tubes cover the zirconia sensor. However, this is in conflict with the aim to accelerate the dynamics of the lambda sensor. In this paper, the heat transfer at the surface of a heated planar zirconia sensor with two different double protection tubes of a Bosch oxygen sensor is examined in detail. The geometric configuration of the tubes forces different flow patterns in the inner protection tube around the zirconia sensor. The zirconia sensor is internally electrically heated by a platinum heater layer.

The success of model-based ECU-functions relies on precise and efficient modeling of the behavior of combustion engines. Due to the limited computing power, usually a combination of physical models and calibration parameters is preferred for engine modeling in ECU. The parameters can be scalars, 1 or 2-dimensional empirical models, such as look-up table for volumetric efficiency and effective area of the exhaust gas recirculation (EGR). A novel algorithm is proposed to automatically calibrate the look-up tables characterizing stationary functional relationships in ECU-function of the air system of a diesel engine with minimum calibration cost. The algorithm runs in the framework of online design of experiment (DoE), in which Gaussian process model (GPM) is adopted to approximate the relationships of interest.