Methoden zur Reduzierung dynamischer Gemischfehler

The improvement of the exhaust emissions of internal combustion engines is one of the most important challenges for the development of software for electronic control units. To operate the catalyst with high efficiency the air/fuel ratio (AFR) has to be exactly controlled in every operation point of the engine. At steady state (constant speed and torque) this can be realized in most cases with a closed loop lambda control. But in transient operation, additionally control functions are required in order to take consideration into the thermodynamic phenomenology of the air path and the fuel path of an IC-Engine. In response model based controllers have recently been developed for the compensation of the main thermodynamic effects in the manifold. However, these controllers are mostly open loop controllers with poor robustness against parameter variations. The main emphasis of this work is the design of new transient AFR-control strategies using linear and nonlinear Kalman-Filters. The realized strategy includes the determination of the in-cylinder air masses, the reduction of the effects from dead times in the fuel path and the compensation of the wall wetting dynamics.
One elementary part is the development of control oriented physical models for the air path and the fuel path. Instead of using only mean value models for the air path, also event based deterministic perturbations (pulsations) were modeled with signal theory techniques. This perturbations are together with the nonlinear filling behavior of the manifold mostly the reason of a speed dependency during parametration of throttle models [Sche-1]. Load transients are able to cause temperature transients in the manifold of more than 30°C. This behavior is modeled and verified with vehicle experiments. The wall wetting dynamics are modeled with a second order lag with feed through term. The two time constants in this model are corresponding to the wall film dynamics and the valve film dynamics. If the engine is not operated at the stochiometric point the balance of controlled and measured air/fuel ratio changes with the rate of residual gas. This effect can easily be used to determine the residual gas ratio in the engine by comparing controlled and measured AFR.
For a speed-density injection system an adaptive Extended Kalman-Filter was used to estimate the air masses in the cylinders. With an additional adaptive parameter for throttle tolerances, the algorithm is not only able to handle strong intake parameter variations but it can also identify air leakage in the manifold. The Kalman-Filter operates in crank angle domain and therefore the frequency of the considered perturbations is kept constant. The algorithm delivers high dynamic manifold pressure and temperature information and the air mass flow in and out of the manifold with small perturbations.
Furthermore the methodology of sensor data fusing is applied to the application of air mass estimation also by using an Extended Kalman-Filter [Sche-3]. If several sensors are available that deliver information about a desired value the sensor data fusing algorithm combines the sensor signals and weights them dependent on a stochastic model. This kind of algorithm delivers very reliable estimates using redundant information.
To reduce the effects of dead times in the fuel path (time between air mass calculation and closing of the intake valve) an air mass predictor was developed.
In a first step this algorithm predicts the effective area of flow by assuming the throttle variation with a colored noise process. In a second step the predicted cylinder air masses are calculated with the results of the throttle predictor by using a free running air path model.
Especially during engine start and at low temperatures the wall wetting dynamics have strong effects on AFR. Open loop dynamic compensators are often realized by using the model matching method. However this controllers have poor robustness in case of variations of the dynamic parameters. With a state space controller and observer the robustness can essentially be increased by using feedback from an Universal Exhaust Oxygen Sensor .