Model based assessment of real-driving emissions – an office and hardware-in-the-loop simulation approach

Johann C.  Wurzenberger, AVL List GmbH

In 2017 the European authorities foresee a gradual introduction of a test procedure for Real
Driving Emissions (RDE). The procedure can be seen as a severe tightening of the emission
regulation since studies show that e.g. the current NOx on-road emissions of modern diesel
passenger car engines are significantly higher than the certified emission limits. The reduction
of NOx is a specific technical challenge. It needs to cope with more demanding (real-life)
engine operating conditions and it also needs to compromise with reduction targets on carbon
dioxide. The fulfillment of both emission limits requires a widening of the focus from an
isolated engine or exhaust aftertreatment view to a system engineering view involving all
hardware and software domains of the vehicle.

Modelling and numerical simulation of complete vehicles is a demanding task per se. It
requires a broad variety of models that describe thermodynamic, chemical, mechanical, fluid
mechanical and electrical phenomena with acceptable predictability, high computational
performance and reasonable effort of parameterization. The virtual assessment of RDE is
additionally challenged by the stochastic nature of the test being influenced by environmental
conditions (temperature, sea level, road type and inclination), driving style and aging
phenomena in various hardware parts. Here, Hardware in the Loop (HiL) environments
constitute a cost efficient and quality increasing methodology as they combine simulation
with hardware prototypes of the control units, sensors, and actuators and therefore is one step
closer to the real system compared to pure office simulation.

This work presents a real-time capable vehicle modelling framework for HiL based
simulation. The key characteristics of the thermodynamic engine combustion and pollutant
formation models and of the exhaust aftertreatment system are discussed on the example of a
modern diesel engine powered passenger car. Here, special attention is paid to three dedicated
areas. First, the scope of the model is tailored to serve as a proper virtual plant. The model
interfaces via the same actuator and sensor channels as they link the real hardware with the
real Engine Control Unit (ECU). This requires, for example, injection and combustion models
sensitive to injection timing signals or models describing the impact of temperature sensors
when feeding back measured temperatures to the ECU. The second area emphasizes the
computational requirements of HiL simulation. HiL capable models need to run significantly
faster than real-time within each single time step (typically given by the update frequency of
the ECU) of an entire simulation duration. This is challenging for implicit and also step-size
adaptive time integration methods that are used to solve the often numerical stiff phenomena
taking place in exhaust aftertreatment devices. An appropriate time scale analysis of the
involved phenomena and a dedicated office simulation assessment of the computational speed
is shown on the example of a diesel exhaust line. This enables a workflow to consistently reuse
not necessarily real time capable office models also in real-time on HiL environments.
The third area presents an approach to mimic real-driving conditions, comprising models of
the vehicle, the driveline, the driver and an approach to randomly compile driving tasks. The
overall simulation framework is suitable to support the assessment of RDE in office and in
HiL simulations.

Download oral presentation: