Challenge X is a three-year program (now in its final year) that tests teams from 17 universities to come up with technology to reduce automotive energy consumption and emissions, and integrate their solutions into a Chevy Equinox SUV/passenger-car crossover vehicle. While previous student competitions focused on hardware mods, the current contest includes a strong modeling and simulation component, as well as subsystem development and testing. Read more about Challenge X and check out the competition's website for the latest developments leading up to the final "drive off" May 30 to June 7, 2007.
Developing an operational automotive hydrogen fuel cell system in three years is a daunting task for any group of engineers. When you take into account a climate that will only let you drive outside for eight months of the year, the task is nearly impossible—without the correct tools that is.
In order to continue to develop our Challenge X vehicle control strategy through the cold winter months in Waterloo, Ontario, our team needed to come up with an accurate vehicle model that could be used to tune controls with software-in-the-loop (SIL) and hardware-in-the-loop (HIL) methodologies.
To ensure model accuracy, empirical models were constructed using actual vehicle data. To parse the gigantic vehicle data logging files that were generated during even short runs and gather data to build the vehicle plant models, a mission-critical .m function was written (details to follow). In this way, model-based design gave us a way tune our control strategy even during a Canadian January.
Model-based design
During our first year of the three-year Challenge X vehicle development, the University of Waterloo Alternative Fuels Team (UWAFT) depended heavily on The MathWorks tool Simulink® vehicle models generated with the U.S. Argonne National Labs Powertrain Systems Analysis Toolkit (PSAT) to select the optimal vehicle powertrain architecture. Once the team had optimized powertrain and fuel selection, it was time to build more in-depth models. We had to incorporate the level of detail into our models that would allow us to perform software-in-the-loop and hardware-in-the-loop control validation, as well as compose accurate vehicle technical specification (VTS) predictions. The figure below shows our modeling approach and makes it clear how heavily the UWAFT team relied on model-based design to reduce development time and improve performance.
Creating models to accurately represent a complex physical system such as a fuel cell SUV required some insight and determination, and powerful modeling tools. SimDriveLine, an add-on to Simulink, made dynamic vehicle performance modeling accessible to a team which was literally working around-the-clock in the shop. The vehicle model developed for VTS predictions is an expansion on PSAT, and is shown in below.
View a full-size image
One of the most valuable additions to the PSAT model was voltage-limit modeling, which incorporated real battery, fuel-cell, and motor performance to optimize regenerative braking. We also updated the PSAT fuel cell model based on data that was collected, filtered, and analyzed in the MATLAB® environment.
