Skip to content

Parametric Validation Real Trips

Jump to bottom
stefanosts edited this page Aug 3, 2016 · 6 revisions

The present section describes the validation of Parametric CO2MPAS using data of real trips measured for a pool of known cars.

For each car several keys characteristics are available: model, fuel, capacity, power, torque, gearbox type, tyres, reported emissions, max velocity, and masses. Additionally, for every trip and for each repetition of the same trip, data are recorded in a frequency of 1 Hz. Among others, the ones used for the Parametric CO2MPAS validation include velocity, engine speed, engine temperature, CO2 emissions, altitude, etc..

In order to respect the sensitivity and confidentiality of the data no data are given here.

#Purpose

This exercise allows the validation of Parametric CO2MPAS "series results" as compared to data measured on the road. More specifically, when providing the velocity profile an assessment is performed on the calculation of the engine speed, the engine temperature and then the fuel consumption / CO2 emissions.

On top of that, the present exercise allows for a first estimation of the validity and potential of Parametric CO2MPAS when used for the estimation of real life CO2 emissions and fuel consumption.

#Methodology

Initially, since not all data required to run Parametric CO2MPAS are available, each car is "located" / identified in the Web Database, based on the following vehicle characteristics: model, capacity, max power, fuel, transmission type, wheel drive and carbody (carbody of vehicle 3 is different that what is found on the web database). Data are then updated with the given test mass as the new reference mass, and used as an input for the Parametric CO2MPAS runs.

The runs are performed for four different "scenarios":

  1. The known engine speed and temperature are provided as an input to the model; that way the error of the tool on the calculation of the engine power and then fuel consumption is assessed
  2. The engine speed and temperature are calculated by the model; the difference between these runs and those of "Scenario 1" provides an estimation of the error of the tool in the calculation of the engine speed and engine temperature and its effect on the fuel consumption and CO2 emissions
  3. The engine speed is provided as an input, while engine temperature is calculated
  4. The engine temperature is provided as an input, while engine speed is calculated.

#Results

Some first results of the Parametric CO2MPAS runs are presented bellow.

Results for one vehicle and one repetition of one trip

Summary results for all repetitions and all trips

The distribution of all trips - and all repetitions - errors

Aggregated results for all vehicles and each trip is given

Discussion - To be further developed

Above we see some aggregate results for the various cars and trips. Overall we have a performance within the expected limits which provides a promising perspective for further applications of CO2MPAS Parametric for real world emissions simulations.

Overall we get an average absolute error of 12.5%, while braking down further the results we have the following:

Per vehicle we have an error of

  • Vehicle 1: 8.96%, StD: 3.75%
  • Vehicle 2: 16.84%, StD: 6.36
  • Vehicle 3: 11.76%, StD: 6.60.

Per trip

  • Trip 1 - Urban: 10.77%, StD: 7.92%
  • Trip 2 - Suburban: 9.68%, StD: 5.71%
  • Trip 3 - Urban: 15.59%, StD: 5.27%
  • Trip 4 - Rural & Motorway: 13.99%, StD: 3.37%.

A more in-depth analysis, evaluating the instantaneous signals, can and will be performed in order to gain a better understanding of the sources of the obrained errors, ways to improve the overall accuracy and decrease the standard deviation of results, etc.




Comparison with reference cases with extra inputs

The results which are presented above concern the case where all outputs are calculated by CO2MPAS Parametric. When additional inputs are provided / used in the simulation, then results change as follows:

  • When engine temperature and engine speed are provided we get an average error of 12.40%
  • When engine speed is provided we get an average error of 12.06%
  • When engine temperature is provided we get an average error of 12.56%.

When additional inputs are used, a small effect is observed on the overall error - as expected - but it could be considered negligible. When both engine temperature and engine speed are provided, sources of inaccuracy should be decreased, thus significantly improve the results. Here, a slight decrease of the mean absolute error is observed but not in the order of magnitude that we would expect. Similar results are observed when only one of the two inputs are provided.

In total, a quick conclusion would be that other sources of innacuracy (calculation of engine power / road loads, innacuracy of real data, etc) have an effect which highly surpasses the effect of the engine speed and engine temperature calculation, thus results are unaffected. Though, in order to reach more solid conclusions about the sources of the above, we need to study in depth the data both in terms of instantaneous signals and individual trips' errors, trying to understand whether there are consistent inaccuracies in the simulation, their sources, etc.

Clone this wiki locally