11. How to assess CO2MPAS results.

This section provides instructions on CO₂MPAS usage so that one can be confident that the tool is used as it is originally planed and with all its potential.

• Make sure that the versions of CO₂MPAS tool and input file are the most recent ones and that the code is coming from the original source.

• Check that the installed libraries of python are the ones distributed with the All-in-One and not updates have been manually done. In order to get a list of the installed libraries and their versions in your Python environment, open a console (Windows Start Menu --> CO2MPAS --> open bash console) and type pip ls > libraries.txt. This command will create a text file containing all the Python installed libraries and versions.

• The CO₂MPAS input file has only the data required for its use.

• Check that all necessary input data have been provided as should. This covers also time series (velocity, engine rpm, engine load, alternator current, battery current, etc.). The CO₂MPAS output_report provides indications by means of the model scores, on the behaviour of the model and potential issues with the input data.

• All input time series must originate from the same measurement (or simulation). Mixing time series between different measurements (eg RPMs from test A with engine temperatures from test B) will most likely lead to failures in the calibration of certain submodels and eventually to inaccurate results. CO₂MPAS calibration is based on a series of methods that correlate the instantaneous performance of different modules. Mixing up of input from different sources should be strictly avoided in case of official (Type Approval related) runs.

• Check that all necessary time series input data are properly aligned (OBD data against dyno data and data measured by other data-loggers). Misalignments of more than 1 second are known to produce inaccuracies particularly in the case of vehicles with automatic transmissions.

• Note that the WLTP test masses in CO₂MPAS input file (vehicle_mass WLTP-H and vehicle_mass WLTP-L) correspond to the masses applied on the dyno during the test and not to the WLTP theoretical ones. For NEDC vehicle inertia (eg. vehicle_mass NEDC-H) the corresponding inertia class should be used as input.

• Engine idle fuel consumption cannot be 0 even for vehicles equipped with Start/Stop technology. The idle engine fuel consumption corresponds to the idle fuel consumption of the vehicle (engine coupled with clutch/torque converter and gearbox) at warm conditions (eg. after one WLTP cycle) when the battery state of charge is at balance conditions and the engine start-stop function is disactivated. Check a tip on how to measure this parameter in the lab.

• Check that the electric currents of the alternator are:
  * not greater than zero (with the exception of engine start events), * zero (+-0.3A) when the engine is switched off * properly time-aligned with the rest of the signals (eg. current is zeroed at the second an engine stop event takes place)

• Check that no NEDC input data (time, velocity, gears) are provided without being necessary. Such data should be provided only in cases where the vehicle cannot perform the full official NEDC (capped cycles) or in case alternative gearshiftings are allowed (eg, starting directly from second gear).

• Check that the first Start Stop event during NEDC occurs as expected. For example if the Start Stop activation time is set to 50sec in the input file and in NEDC, the engine switches off during idling phases much later than 50sec, this will most likely cause errors in the prediction of the model.

• In case of automatic vehicles check that predicted engine RPMs during UDC and EUDC plateaus are not lower than those actually measured. If lower than those actually measured repeat the simulation with the input variable (fuel saving A/T strategy set to 0 and compare which of the two RPM predictions returns the lowest error (as reported in the output_report) compared to the measured RPM signal.

• A series of technology options are enabled in CO₂MPAS. Some of them are at the moment -Feb. 2017- not fully validated due to the lack of necessary test data. They can be used for improving CO₂MPAS accuracy. It is advisable that simulations are run both with and without these technologies and that the configuration that gives the best accuracy is used. In cases of random-testing however it should be verified that the configuration that has been actually used led to more accurate CO₂MPAS results and not to lower CO₂ values.

Back to the Index of the user guidelines