This dissertation studies improvements in the application of model-based design for engine and powertrain control systems in automotive vehicles. The focus of study is to explore the availability of the model-based design to the control of dynamic systems that are hard to model, as in the case of engines and powertrain components.
The study has two thrusts that subsequently spawn a novel perspective that possibly enhances the traditional model-based design to a new level.
First, a new powertrain control method is proposed which highlights the usage of the model-based design to achieve a control objective more systematically than conventional methods. The objective here is to achieve shockless gear shifting under conventional hardware conditions. The method utilizes a collaboration between engine and transmission, a concept which has not been exploited extensively in the conventional design. Through the model-based design based on the modeling of the transmission components, the method provides a systematic process which will generate detailed control setpoints for engine control. Availability of detailed engine torque control is an important key in improving powertrain control performance, due to the engine's nature as the powerhouse of the whole powertrain.
As the second thrust of this study, a new method is proposed that effectively bridges the gap between theoretical modeling and practical reality of the powertrain systems. The subject of study here is robust engine control. The objective is to accurately control generated torque and air-to-fuel ratio (AFR), both of which are important engine outputs by which the performance of the engine is measured. Robust control based on model-based design assumes a nominal model of the engine. To compensate for the model discrepancy between the nominal model and the actual engine dynamics, disturbance observer (DOB) techniques are applied in this study. To further this model-based design based on DOB, an adaptation of iterative learning control (ILC) is introduced to enhance the robustness of the control scheme. The result is a new control scheme that is a structured combination of DOB and ILC. This scheme actually has a vast potential as a new type of robust control implementation technique that is both effective and easy to implement in many control applications extending out of automotive powertrain control. Numerical and experimental evaluations under various conditions show the effectiveness of the proposed schemes based on DOB and ILC.
Finally, in perspective, it is seen that beneath the above two thrusts of this study underlies the traditional model-based design which is based on the initial assumption of the nominal plant model obtained by off-line modeling of the plant. Then, it will also be discussed that the model-based design is actually being reinforced and enhanced by incorporating the learning and compensation of model uncertainty that is encountered only through on-line operations. In other words, the model-based design approach lends itself to a data-driven, data-rich approach that utilizes the knowledge of model discrepancy learned by ILC, in the form of a signal data sequence, to supplement the nominal plant characteristics.
Based on the observation above, a novel modeling framework, although unverified, is proposed in order to model a plant with substantially increased accuracy. Although still based on the assumption of a nominal model, the framework aims to utilize the model discrepancy information in an organized manner. The modeling framework may, so to speak, model the un-modeled.
The keynote throughout this dissertation is to provide new measures to help promote the application of the model-based design, especially where the model-based design has not been exploited in conventional designs and applications. Model-based design is a universal approach that can be applied to any kind of control system, so is its extension elaborated in this dissertation.