Heavy-Duty Diesel Engines

Diesel engines continue to be the prime movers in heavy-duty and nonroad applications because of their high fuel efficiency and excellent durability. Optimising and controlling the engine and its exhaust-gas aftertreatment system (ATS) poses a challenging task. On the one hand, low fuel consumption and low pollutant emissions, in particular nitrogen oxides (NOx), are conflicting objectives. On the other hand, the system's complexity and the number of control inputs are high. In this research project, we tackle three challenges related to the optimal control of Diesel engines.

There are several control inputs that influence the engine's efficiency and pollutant emissions, such as the injection timing, the rail pressure, and the amount of exhaust-gas recirculation (EGR). In order to achieve the highest possible efficiency while limiting the engine-out emissions, the engine control strategy needs to be carefully tuned. This is known as engine calibration and has been extensively studied since the application of electronic control.

When calibrating the engine, the two main challenges are the large number of degrees of freedom and the trade-off between fuel consumption and NOx emissions. Thus, the calibration task consists of a multi-objective optimisation problem with conflicting objectives. This means there is no single best solution, but rather a set of Pareto optimal solutions. Two research questions are therefore:

• How can we guarantee an optimal engine calibration?
• Which point on the optimal trade-off curve should we choose?

The development and calibration of the engine and of the ATS are typically conducted separately, using individual targets largely based on engineering experience. In order to exploit the full potential of the overall system, as well as to streamline the calibration process, a so-called supervisory control strategy can be used. Its key element is the use of a variable engine calibration, which allows the engine-out NOx emissions to be changed in real time depending on the performance of the ATS. To a certain extent, this idea has already been put into practice in the industry by using several engine modes.

In a more systematic approach, the combined system is optimised by formulating and solving an optimal control problem (OCP). Supervisory control strategies developed so far require an optimisation problem to be solved online on the electronic control unit (ECU). Considering the complexity of the models involved, the number of degrees of freedom, and the limited computational resources of industrial ECUs, this can be a major hurdle to technology adoption. Therefore, the following research question is posed:

• How can we design a supervisory controller that does not require an optimisation problem to be solved online, is easy to implement and tune, yet achieves near-optimal performance?

Until the 2010s, the test procedures for the type approval of engine systems were largely based on standardised test cycles. As a consequence, emissions measured on the road have often signifficantly exceeded the legislative limits. In order to reduce the gap between type-approval and real-world emissions, real driving emissions (RDE) testing has become an important legislative tool.

The core component of the RDE legislation is the so-called moving averaging window (MAW) method. Its purpose is to limit the pollutant emissions throughout the in-use operation, rather than limiting only the average over a (standardised) test cycle. This is achieved by defining many overlapping time windows and limiting the emissions accumulated during each window. In the literature, these so-called window constraints have not been explicitly included in the OCP so far. Rather, most approaches use a robust control strategy and thus sacrifice performance. Therefore, we pose the following research questions:

• What are the expected gains of an optimal RDE control strategy compared to a robust one?
• What is the impact of the windows of the MAW method on the optimal control policy?

S. van Dooren, C. Balerna, M. Salazar, A. Amstutz, and C. H. Onder. external pageOptimal Diesel engine calibration using convex modelling of Pareto frontierscall_made. Control Engineering Practice, 96:104313, 2020

S. van Dooren, A. Amstutz, and C. H. Onder. external pageA causal supervisory control strategy for optimal control of a heavy-duty Diesel engine with SCR aftertreatmentcall_made. Control Engineering Practice, 119:104982, 2022