(322c) Design and Application of Predictive Functional Control On An Automotive Catalytic Converter | AIChE

(322c) Design and Application of Predictive Functional Control On An Automotive Catalytic Converter


Schallock, R. W. - Presenter, Villanova University
Muske, K. R. - Presenter, Villanova University
Peyton Jones, J. C. - Presenter, Villanova University

In this work, a model-based predictive functional controller for automotive three-way catalyst system control is described and implemented on an experimental Ford 2.0 liter I4 Duratec spark ignition engine. Model-based control techniques offer an attractive advanced control methodology for automotive three-way catalyst systems. The predictive properties of model-based control, which has led to its widespread adoption in industrial applications, are ideal for preventing emissions breakthrough before they occur. Because model predictive controllers are based on the dynamic optimization of a specified cost function, they offer the ability to consider competing control objectives, such as emission reduction, performance, and fuel economy, in a systematic manner.

There are a number of requirements for the on-line implementation of model-based control that must be considered in automotive applications. The most critical of these requirements is a model-based control algorithm that is computationally efficient in order to execute within the limitations of a standard engine control computer. The key aspect to meeting this requirement is an efficient dynamic on-line optimization methodology for the specified performance objective. Related requirements are a catalyst model that can adequately predict the dynamic behavior of the three-way catalyst system and the ability to estimate the current state of the model from available measurements with modest computational effort. Finally, a base control system that provides acceptable regulatory and servo response for the commands issued by the model predictive controller is also required. These requirements are addressed in this work through the implementation of predictive functional control in a multi-rate cascade control configuration.

The catalyst controller is a nonlinear model predictive controller that incorporates a dynamic catalyst model to determine the post-catalyst exhaust gas oxygen (EGO) sensor relay controller parameters required to minimize the desired catalyst operating objective. The simulated performance objective presented in [1] that tracks a desired catalyst stored oxygen value is considered in the experimental application in this work.

The predictive functional controller in this application essentially finds the optimal post-catalyst EGO sensor relay controller for the specified performance objective. It is referred to as a predictive functional controller because the functional representation of the pre-catalyst operating trajectories that are considered by the optimization is characterized by the set of relay feedback controllers obtained through adjustment of the relay controller parameters. In the experimental demonstration of this controller presented in this work, the EGO sensor relay target is adjusted by the predictive controller. The result is a one-dimensional optimization problem that can quickly be computed on-line using a variant of Brent's method. This cascade structure can also tolerate failure of the catalyst controller to solve the on-line optimization problem at every control interval because the post-catalyst EGO relay controller will continue to stabilize the catalyst system.