Anna Stefanopoulou

Current Projects

Collaborations

 Ford Motor, General Motors, United Technologies, Mack Trucks, Turbodyne Systems

Funding: NSF,  DOE,  PATH, ARC

Over the last decade there have been dramatic improvements in proton exchange membrane (PEM) fuel cells that enabled fuel cell power to transcend from the laboratory to experimental vehicles. However, the viability, efficiency, and robustness of this technology depend on understanding, predicting, and controlling the unique transient behavior of the Fuel Cell (FC) breathing system. Although steady-state FC behavior is considered the normal operating mode; start-up, shut-down, and sudden load changes are characteristic and ubiquitous to all power producing devices. During all operating modes, our ability to precisely control the reactant flow and pressure, stack temperature, and membrane humidity is critical.

To this end, phenomenological models and robust control methodologies are developed to address the subsystem conflicts and account for the nonlinear interactions and constraints imposed by sensor fidelity and actuator authority. Insight and rigorous metrics are provided for the vehicle power management and level of hybridization with battery and/or ultra-capacitor. Finally, the impact of the control architecture for the coordination of all the three electric power sources (FC, battery, ultra-capacitor) with the traction motor inverter is systematically analyzed.

This project allows students to develop control theoretic tools for the highly interdisciplinary areas of FC vehicles and power systems. Both of these technologies are important to our national competitiveness and a sustainable environment. (in collaboration with Scott Bortoff, UTRC, funding from NSF)



Through this project we integrate instrumentation and equipment for the development of Control and Diagnostic Systems for Fuel Cell Power. The experimental set-up allows the implementation of multivariable controllers, fault detection, and diagnostic algorithms for the regulation of reactant flow and pressure, stack temperature, and membrane humidity. It is anticipated that the development and testing of real-time control and diagnostic systems will accelerate the use of Fuel Cells by enhancing their safety, increasing their efficiency, and ensuring their robustness in real world applications.

The equipment and control testing facility augments the University of Michigan capability in FC research. The university's strategic location and collaborations with the automotive industry generates synergistic mechanisms for research and for training the future technical and corporate leaders in the field. (funding from NSF, collaboration with SERC )

Picture from the 1082 Auto Lab August 2002

  • Coordination of hydrogen and air flow for transient Fuel Cell loading

    • Our objective is to develop the analytical framework and the methodology for calibrating the multi-loop Proton Exchange Membrane Fuel Cell system for high efficiency during transients typical in urban driving cycles. Different sensor/actuator characteristics will be considered in order to determine best and worst case scenarios. To achieve these goals we use physical principles and empirical relations to develop the dynamic model, and experimental vehicle data for the model validation (whenever possible). We employ multivariable control analysis and synthesis techniques for defining the system architecture and calibration.  This work wil  be  instrumental in developing FC-hybrid army vehicles and robots for unmanned operations.  The army is currently evaluating FC for use as auxiliary power supplies to support "Digitization of the Battlefield." Our work will provide modeling, analysis, and simulation  tools to evaluate PEM-FC performance during high bandwidth loads typical in real world conditions. (Funding from US. ARMY, TACOM)

    "Modeling and Control of Fuel Cell PEM Stack Systems ,"  in 2002 American Control Conference Proceedings, May 2002.



    Homogeneous Charge Compression Ignition (HCCI) combustion achieves high fuel efficiency with low emissions without any major hardware modification of traditional internal combustion engines used today in automotive vehicles. The main difficulty in the HCCI combustion isthat we cannot control directly the combustion process by the spark timing nor the fuel injection timing that are used to initiatecombustion in Otto and Diesel engine, respectively. The combustion isindirectly controlled by the mixture properties (temperature,pressure, and composition) and the breathing sequence (intake, compression, and exhaust timing). Although several studies have been performed in automotive research laboratories to determine which parameters affect the HCCI combustion, there has been very little work on the development of real-time feedback control algorithms for HCCI combustion.

    To this end we develop multivariable and learning control methodologies in designing a robust feedback controller for variable valve timing (VVT) and spark discharge energy (SED). This will be accomplished by designing a tracking controller that involves a low order feedback using SED and a cycle-to-cycle iterative learning controller (ILC) that generates valve trajectories that achieve the desired combustion boundary conditions, and thus, the optimum HCCI breathing process. The ILC controller learns the input/output dynamics and then it predicts and schedules the inputs that can best affect the output histories. This approach is suited for quasi-periodic processes with uncertain and nonlinear dynamics between the inputs and the outputs. (funding from GM)

    The extra degrees of freedom obtained from a camless valvetrain constitute the beauty and the curse for the camless engine control system. Indeed, the flexibility of controlling the intake and exhaust valve timing and duration can alleviate many otherwise necessary engine design tradeoffs. Specifically, it has been shown that controlling the intake valve events can eliminate the need for throttled operation in gasoline engines, thereby reducing pumping losses and improving fuel economy. Other benefits of camless engines include higher torque output, cylinder de-activation, and elimination of external exhaust gas recirculation. Camless valvetrains are under intensive development by several manufacturers, with the Electromechanical Valvetrain (EVA) currently considered by many to be in a relatively more developed stage.

    We develop and experimentally demonstrate a robust control system for a gasoline camless engine with an EVA actuator in a dynamometer test cell at the University of Michigan. Our work concentrates on two critical problems, namely, the soft landing of the valves and electromechanical valve actuators (EVA), and the transient air and residual management.  The first problem directly addresses durability and noise issues associated with the electromechanical valve actuators; the second addresses the complexity issues associated with the transient control of the extra degrees of freedom with a minimum set of existing of vehicle sensors. (funding from NSF, DOE, Ford)

  • Multivariable Controller Architecture for Advanced Powerplants

    • The global need for improved fuel economy and reduced emissions often requires innovative mechanical engine configurations (new actuators) that introduce additional design parameters (control variables) used to optimize the engine performance. These new actuators result in highly complex powertrains with significant coupling between subsystems thus posing challenging multivariable control problems. In this work we address the fundamental question of defining a low complexity controller architecture for these advanced technology powerplants (funding NSF}.
     

  • Diesel Technology

    • We formulate the control problem that addresses emissions and engine performance requirements of a diesel engine equipped with variable geometry turbocharger (VGT) and exhaust gas recirculation (EGR). At the optimal points of engine operation the multivariable system has an inherent limitation: locally, the two available actuators collapse to a single degree of freedom. To address these difficulties we employ multivariable feedback theoretical concepts pertinent to singular value decomposition and design a feedback controller that ensures coordination of the two actuators in response to feasible tracking commands {in collaboration with I. Kolmanovsky, Ford Motor Company}
     

  • Advanced Braking Methods for Longitudinal Control of Commercial Heavy Vehicles (ABC4CHV)

    • In this project we develop models, analysis and the control methodology for integration of advanced braking methods with conventional braking systems for longitudinal control of commercial heavy vehicles. Over the last ten years there has been a significant improvement of the reliability and efficiency of the heavy duty vehicle powertrain. Nowadays, heavy duty vehicle acceleration and headway velocity is comparable with passenger vehicles. This transformation was achieved primarily by using lightweight material and components and by reducing the aerodynamic drag and frictional losses. Increase in operational vehicle speed combined with decrease of the natural retarding capabilities in modern powertrains creates challenging braking requirements. Compression braking is a retarding mechanism that combines high braking capability with fuel economy benefits and potential emission reduction. It is achieved by inhibiting fuel injection, altering the conventional gas exchange process and transforming the engine to a giant compressor that absorbs power. We develop control algorithms and specifications for anti-lock braking, automatic traction control and cruise control in light of the additional braking source (in collaboration with Joe Schmidt, Mack Trucks, funding PATH ).
     

  • Modeling, Analysis, and Control of Flow-Assist Mechanisms

    • Novel flow-assist mechanisms such as an electric turbocharger or an electric supercharger can potentially reduce turbolag and smoke during acceleration, increase mixture dilution, and improve scavenging in 2-stroke engines. To realize the potential benefits of the additional actuators without detrimental consequences to the fuel economy we analyze their dynamic interactions with the intake and exhaust engine process and develop calibration procedures using nonlinear multivariable control theory. Our analysis will be based on a mean value engine model. Empirical emission models based on neural networks will be included in our engine model to assess performance tradeoffs between fuel economy, emissions, and drivability (in collaboration with Turbodyne Systems Inc.)
     

  • Modeling and Control of IC Engine with Variable Valve Motion

    • Availability of fully variable camless actuation presents a great opportunity for substantial improvements in engine operation as well as a great challenge in being able to cope with and properly use the many new degrees of freedom that become available for engine optimization. This project requires integration of various aspects of fluid, thermodynamic and systems theory for the development and the control design of the variable valve motion engine. The developed model is intended to fill the gap between the analytical thermodynamic engine models and the steady-state engine models that are currently used for studying camless engines. In the controller development we address observability issues arising from cylinder-to-cylinder overlap. Estimation and controller adaptation is deemed to be necessary due to the high parameter variation and model uncertainty (in collaboration with Mike Levin,  Ford Motor Company).

    Anna Stefanopoulou Home Publication List
    [Mechanical Engineering ] [College of Engineering ] [UMICH ]
    Last Updated: May, 2002 -AgS