Physics-based controller design to reduce data dependencies and engine calibration workload 2018-36-0170
Model-based controls development has the potential to reduce cost and development time and helps to improve quality of the control design. It helps creating an infrastructure more robust to unforeseen changes to the development cycle plan. More upfront effort is required, including physical parameter characterization, model development, and “potentially” increased demands for processing power and memory, but the longer term benefits outweigh these costs. On top of that, the language of physics can be used to characterize a system by reducing its behavior to a set of differential equations. The main benefit is a potentially significant reduction in calibration effort, since there will be less calibratable parameters, tables and logic operators. Calibrations are reduced to dynamic system parameterization, thus reducing dependencies on empirical data during the control design. This also allows for future fidelity improvements without tearing up the architecture. These benefits find a roadblock in hardware constraints on embedded applications, where increased complexity of the dynamic equations might affect real time capabilities. This paper explores these benefits in the development of a software module comparing the calibration effort on both the physics based and the empirical data based applications. The particular system under investigation is the embedded physics model for boil-off phenomenon that occur in internal combustion engines (ICEs) in flex-fuel vehicles (FFV) applications, used during the compensation of injected fuel. The ethanol vaporization from the oil sump was parametrized using physics modeling, facilitating its development and understanding. The model for fuel drainage to the sump is out of the scope for this paper.