The Hybrid Modeling Technique – Enabler for Adaptive and Real-Time Capable Process Models of High Accuracy for Future Powertrain Control Concepts 2019-01-1178
Constantly increasing demands on the powertrains of modern vehicles require the optimization of conventional combustion engines in combination with tailored electrification and vehicle connectivity strategies. The resulting powertrain systems feature a high degree of freedom due to the large number of available adjustment parameters. This obviously presents major challenges to the development of the corresponding powertrain control logics. Hence, the identification of an optimal system calibration is a non-trivial task. To address this situation, physics-based control approaches are evolving and successively replacing conventional map-based control strategies in order to grasp more complex powertrain topologies. Physics-based control approaches enable a significant reduction in calibration effort and also improve the control robustness. However, due to the requirement of real-time capability, physical models have to be formulated via simplified mean value approaches, which in turn limits the control accuracy.
To eliminate the constraints of a purely physics-based control approach, the underlying physical process model can be extended by an additional data-driven model component. For this purpose, the artificial neural network (ANN) is considered. ANNs are well-known for their high model accuracy, but they show a poor predictive robustness in regions which have not extensively been trained with data beforehand. Consequently, a new modeling strategy can be derived: The general process tendencies are estimated by a physical model, while the ANN corrects the estimation in well-known regions to maximize the overall process model accuracy. This novel approach is referred to as “Hybrid Modeling Technique” (HMT). It follows the general idea of combining advantageous attributes of a physical model (predictive robustness and low calibration effort) with the benefits of an ANN (local model accuracy). Besides the achievement of an improved process model accuracy, HMT further enables model adaptations by involving a dynamic ANN training, which can be repeated when the process behavior changes as a consequence of e.g. hardware drifts, aging or different process boundary conditions.
To evaluate the performance of the given HMT, the approach is exemplarily applied in this contribution to derive an adaptive, real-time capable Diesel ignition delay model and an engine-out NOx emission model. In both cases, an available physics-based model is used as a baseline. In accordance to HMT, the physics-based model is extended by an additional ANN term, which is correspondingly trained to improve the model accuracy in standard conditions, while the physics-based model is used in operating conditions that have not been considered during model training.
Christian Joerg, Sung-Yong Lee, Christoph Reuber, Joschka Schaub, Matthias Koetter, Silja Thewes, Ronnie Thattaradiyil, Jakob Andert