Adjoint-based system identification and feedforward control optimization in automotive powertrain subsystems

Tuning the response of automotive powertrain subsystems can both improve the performance and ease overall system integration in both conventional powertrains and innovative new powertrain designs. Existing control strategies for both the powertrain and its subsystems usually incorporate extensive feedforward maps because the systems of interest are nonlinear and difficult to model, operating ranges of interest are large, and maps are relatively inexpensive to implement. Augmenting or replacing such maps with dynamic feedforward control strategies based on predictive dynamic models can possibly provide significant performance improvements. The present work develops adjoint-based optimization approaches to: 1) identify the several unknown parameters of a nonlinear model of a powertrain torque converter and transmission using empirical data; and 2) based on this identified model, optimize the feedforward controls for a powertrain gear shift to achieve the desired shift characteristics. It is shown that the adjoint-based system identification procedure yields an accurate system model and that the adjoint-based control optimization procedure improves the anticipated system performance.