Presentation is loading. Please wait.

Presentation is loading. Please wait.

Meeting of the Steering Group for Simulation Issues from the last SGS meeting in LAT model development Munich, 25 July 2014 Biagio Ciuffo Georgios.

Similar presentations


Presentation on theme: "Meeting of the Steering Group for Simulation Issues from the last SGS meeting in LAT model development Munich, 25 July 2014 Biagio Ciuffo Georgios."— Presentation transcript:

1 Meeting of the Steering Group for Simulation Issues from the last SGS meeting in LAT model development Munich, 25 July 2014 Biagio Ciuffo Georgios Fontaras Stefanos Tsiakmakis Alessandro Marotta JRC

2 Background “Correlation Function” or “Meta-Model” is the tool chosen in this project for evaluating WLTP-based CO2 emissions from NEDC-based ones and vice-versa given the technical features of a given vehicle. Different approaches can be followed, with theoretical foundations, different levels of sophistication, different inputs, different accuracy and robustness expectations.

3 Meta-model loop (Current)
Real vehicle Validation Calibration Simulation model LAT (+JRC) TUG/JRC TUG/JRC Meta-model

4 LAT study The meta-model development was planned to be based on the results of a number of vehicle simulations The development of proper simulation models is a key step in the entire project as the accuracy of the final meta-model will also depends upon the accuracy of the simulation results Due to its well-known experience, LAT was selected by the TWG to deal with the development of the different vehicle models using AVL Cruise as simulation tool

5 Modeling approach (theoretical)
Vehicle fleet Fleet segmentation and technologies Identification of representative vehicles Simulation model Vehicle test Getting vehicles from the road

6 Modeling approach (actual)
Vehicle fleet Fleet segmentation and technologies Simulation model Identification of representative vehicles Input data provided by OEMs Vehicle test Vehicle provided by OEMs

7 Implications OEMs’ support was not seen in a positive way by some MSs and associations OEMs contribution was very important to get the necessary technologies inside the study and to achieve the required accuracy (crucial to have input data) OEMs could take advantage of the control on the vehicles from which the models have been developed (data from VW Golf, BMW 116i, Audi A6, AR Giulietta, etc. not provided or delayed because the tests not in line with OEMs expectations)

8 Issues in LAT modeling During the last SGS meeting, LAT presented a summary of the modeling activities for three vehicles (Astra, X1, and Auris). A few issues were highlighted (reason to organize this meeting): OEMs’ requests-related issues (need to meet TA values) Modeling-approach related issues (need to follow different strategies for different vehicles and to adjust the vehicle model to the different cycles) ToR accuracy criteria-related issues (need to amend them) Need to urgently deal with them!

9 TA CO2 emissions TA works like this:
An OEM brings a vehicle to the TA authority declaring a CO2 emission X The TA authority measures the vehicle and get a value Y If Y≤X×1.04 the CO2 emissions assigned to the vehicle is X In addition, the vehicle might also have taken the CO2 value extending the figure from a “similar” vehicle emitting up to 4% less

10 TA CO2 emissions What it has been measured at LAT/TUG/JRC can have a significant difference from the TA value up to 4% can be due to TA procedure 2-3% (optimistically) due to lab differences additional x% due to random vehicle differences and test fluctuations Why should we ever take into consideration TA values in the correlation project? In this attempt, it can bring strong distortions into the model TA value should never appear in our discussions!

11 Modeling approach In order to have a physically robust vehicle model, it has to be fixed no matter the cycle that it has to run Given the available data, from a conceptual point of view it should be calibrated on a certain cycle and then validated on one or more other cycles In calibration the model parameters are adjusted until the model reproduces the real data In validation the model is just run as it is after the calibration and the distance between outputs and real data is checked If the validation is not satisfactory, the model must be re-calibrated and re-validated (validation loop)

12 Modeling approach In our project we have at least 3 sets of data:
NEDC data (cold+hot) WLTP data (cold+hot) ERMES/ARTEMIS data (hot) Per each test we have 3 repetitions. They are required to see whether the vehicle behaves in the correct way. Not all of them should be used for the calibration/validation of the model A single repetition of each test should be chosen and model calibration/validation should focus on it A possible approach might be to calibrate the model on the WLTP and then validate it on NEDC and ERMES/ARTEMIS

13 Modeling approach Each OEM has its on strategy/logic in developing the vehicle model The need to agree on the approach to follow with each of them has been proven to be not-efficient and effective JRC proposes to agree on a modeling template instead and to leave LAT free to work following it. LAT is then judged on the final results (ToR accuracy criteria) without intermediate discussions with OEMs if not strictly necessary A first model-template has been developed based on the lessons learned with the model of the Fiat Punto. It contains all the models that are necessary to achieve the accuracy required by the project

14 ToR accuracy-related issues
All the accuracy criteria set in the ToR must be maintained to ensure that the model is properly developed They should not be checked on any test carried out (some of them might be wrong, reason why we have three repetitions) but only on those chosen as reference LAT has really tried to achieve an almost impossible accuracy (considering all the uncertainties of a vehicle test) ToR criteria n. 3 to be amended: Consistency throughout the test. Over at least [90%] of the cycle duration the cumulative fuel consumption calculated on the validation cycle shall remain within [±3.5%], compared to the measurements

15 ToR accuracy criterion n.3
Possible amendment (from meeting at LAT) Over at least [90%] of the cycle duration the instantaneous fuel consumption calculated on the validation cycle shall remain within [±Xg], compared to the measurements. To account for diffusive phenomena, the Ys moving average of the fuel consumption is considered as reference for both model outputs and test measurements X and Y to be defined. Any suggestions? 3g and 5s?

16 Additional issue: Bag vs. Modal CO2
From the LAT presentation at the last SGS it seemed that there is a huge discrepancy between bag and modal CO2 values. How can one achieve the necessary accuracy in bag values (the most accurate) if one needs to follow the modal value to properly calibrate the model? How can the huge discrepancy motivated? What to do to deal with it?

17 Next presentations The JRC proposal for a general template for building a robust simulation model in Cruise


Download ppt "Meeting of the Steering Group for Simulation Issues from the last SGS meeting in LAT model development Munich, 25 July 2014 Biagio Ciuffo Georgios."

Similar presentations


Ads by Google