Model-Assisted Calibration Meets Future RDE Requirements

A virtual vehicle on a virtual test bench permits robust calibration in a short time

When real driving emissions (RDE) are crucial to approving vehicles in less than two years, the work of the calibrators will become even more challenging: Then they must ensure robust calibration that reliably meets the limit values for different traffic conditions, journeys and driving styles. IAV has developed a process for this that is based on modeling the whole vehicle, delivers precise, reproducible results and noticeably speeds up the development process.

Even if the same driver always sits behind the steering wheel, the results of real-world measurements in road traffic with PEMS (Portable Emission Measurement Systems) equipment fluctuate widely. “This makes it impossible to reproduce results exactly or precisely determine how modified control unit calibration affects exhaust emission values”, explains Dirk Neumann, head of the Diesel Engine Calibration / Combustion team at IAV. “In close cooperation among five sections, a process has been developed at IAV that makes it possible to calibrate control units in a virtual environment. We are using this process more and more in customer projects too.”

Models are used for the whole vehicle. Engine emissions are simulated as a dynamic DoE model. It is based on dynamometer measurements for which the engine is dynamically excited in various zones – either sinusoidally or by APRBS (Amplitude-Modulated Pseudo-Random Binary Signal). “This approach is completely data-driven and training is based on signals that can be varied over time”, Neumann says. “The system response from the model equates exactly to the response from the real-life engine.” Among the outputs it provides, it delivers the untreated emissions of NOx, CO, CO2, HC, O2 and soot whereby, in particular, the values for nitrogen oxides and carbon dioxide are very precise and hardly differ from the values measured. The algorithms of the control unit (ECU) and air path (from IAV’s TRSim Air Path model library) are also modeled in MATLAB / Simulink and, together with the dynamic DoE model, form the model of the entire engine.

Physical-chemical model of exhaust gas aftertreatment

The model of the exhaust gas aftertreatment system is generated in axisuite and uses the simulated engine-out emissions and temperature values from the engine model as its input variables. It simulates the physicalchemical processes in the catalytic converters (one to three-dimensionally), with an IAV syngas test bench being used to calibrate the reaction kinetics of components. “The calibrated model is then validated on an engine test bench or a roller dynamometer”, Neumann reports. “It is precise and robust, and also takes account of secondary emissions, such as ammonia.” Here, too, the computed values correlate very well with those measured. Part of the model for the exhaust gas aftertreatment system is the dosing control unit (DCU) for AdBlue which is also simulated in MATLAB/Simulink.

The vehicle’s longitudinal dynamics are simulated with IAV’s VeLoDyn tool which needs input parameters such as vehicle weight, transmission data and powertrain efficiency. Although these can be ascertained on the road or testing ground, the manufacturers can in many cases provide the data that are needed for this. Simulating the driver also forms part of the longitudinal dynamics model with controller parameterization that can simulate various driving styles (sporty, normal, defensive), extending the options in generating a robust calibration for engine and exhaust gas aftertreatment.

Using familiar tools like INCA

For calibration purposes, IAV’s experts put the virtual vehicle on a virtual test bench to simulate various driving styles (standardized cycles or real-world RDE runs under different ambient conditions). To do this, the calibrators can still use the tools they are familiar with, like INCA. IAV’s MiL-Desk (Model in the Loop on your Desktop) tool provides the basis for connecting the models and compiling them in executable files that can be directly linked with INCA. “This not only makes the calibrators’ job easier but also permits very fast computations”, Neumann says. “We only take about a fifth of time needed for real measurements.”

The new RDE calibration process can be used throughout the vehicle development cycle, from advance engineering to the start of production. Depending on circumstances (e.g. the availability of real-world components, such as engine or vehicle), the user can decide whether to calibrate on a model or experimental basis. For a test run, Neumann and his colleagues have provisionally calibrated a production EU6b diesel vehicle using test bench measurements and PEMS drives. They then took their models as the basis for separately optimizing ECU and DCU. Afterwards, the best compromise between AdBlue consumption and CO2 emissions was determined at a prescribed NOx limit and the robustness of calibration assessed. Final validation then took place using PEMS equipment on the road again, with the results from simulation once more correlating very well with real-world measurements.

Successful use in customer projects

In the meantime, the model-based approach for RDE calibration has also proven very successful in initial customer projects. “The process is reliable and we are highly satisfied with the results”, says Neumann, summing up. “In future, we want to team up and take it forward – for instance, by using our new high-altitude climate roller dynamometer in Berlin to make even better allowance for the influence of altitude and low temperatures on calibration. A further emphasis will also be on automatically optimizing tailpipe emissions.