For some time, IAV engineers have been successfully applying dynamic-modeling techniques to engine calibration. Recent advances have produced time-saving methods for improving the dynamic engine-operation phases identified as causing both high fuel consumption and pollutant emissions. Tighter legislation and cost pressures demand the immediate application of intelligent modeling and optimization techniques in the product-development process.
Dynamic modeling considers the time-varying nature of I/O values in a system. As with steady-state methods, optimized dynamic models can extract multiple solutions from the same set of data, while the ability to model transient engine behavior drives the systematic method-development for transient engine calibration. Though parallels do exist between steady-state and dynamic methods, the dynamic approach is more complex and requires expansion.
In expanding the experiment design, the engine operating boundaries must first be determined. With Dynamic DoE, the boundaries vary over time and many combinations of input parameters are addressed. For the best model quality, while keeping measurement costs to a minimum, IAV excites the system with multi-input “chirps”, or APRBSs. These fully dynamic experiments take into consideration amplitude distribution, frequency distribution, engine-operating boundaries, and more.
Dynamic DoE increases test cell measurement complexity and the quantity of data produced. Test automation is a requirement with a strong coupling to the engine controller, and depending on the complexity of the calibration task, vehicle simulations may be made. The simulations can be as simple as the basic duplication of engine running conditions or more complex vehicle models for simulating emission drive cycles.
Dynamic test results gather much time or event-based data from many devices in the test cell. IAV has developed tools to synchronize these data inputs to a consistent data rate.
The simulation of transient engine behavior is based on nonlinear dynamic models from which IAV develops data-driven models, such as Volterra series, Neural Networks, and Probabilistic Models. Feedback terms may be added for the model-output values from an earlier time step may be used to predict the value for the next.
With the models created, the optimization can begin. The objective of dynamic optimization is the same as that of steady-state: minimization of emissions and fuel consumption within the constraints of engine roughness and temperature. The result of the Dynamic DoE optimization is a time-varying trace of optimized inputs rather than a single value.
Dynamic DoE engine optimization considers the impact of current engine operating values on predicted engine responses. Such dependencies normally demand significant computing power, but IAV provides proven techniques for reducing computations to yield the optimization relatively quickly. The Dynamic DoE tools calculate optimum parameter time traces for engine operation, which can be analyzed in combination with a catalytic converter simulation. The time traces can be used, in turn, for deriving new control strategies or for improving fuel consumption and emissions.
IAV supports its customers by helping to integrate Dynamic DoE tools into the process chain, by supplying documentation and training, and by offering assistance in developing new fields for the application.