The state-of-play of the NEDC/WLTP Correlation project

Transcription

1 Meeting of the Technical Working Group on Correlation (TWG) The state-of-play of the NEDC/WLTP Correlation project Brussels 27 November 2014 Disclaimer: The views expressed are purely those of the writer and may not in any circumstance be regarded as stating an official position of the European Commission Biagio Ciuffo Georgios Fontaras Alessandro Marotta Stefanos Tsiakmakis Kostis Anagnostopoulos JRC

2 The NEDC/WLTP correlation project Since its first meeting aim, scope and complexity of the project have considerably evolved OEMs and MSs have pushed towards unexpected levels of accuracy to assess the comparable stringency criterion set by the study From fleet-wide to vehicle-oriented approach 2

3 Effect of the evolution Need to predict CO 2 differences between the two test procedure for any vehicle up to 2020 (or more) with high accuracy The time plan required to be considerably adapted (from a few months to possibly several years with the need to check the accuracy as technologies evolve) Additional projects from MSs were necessary for new tests/models The project has evolved towards un unexpected level of complexity having almost all main OEMs and research groups working in this field involved (JRC, LAT, TUG, Ricardo-AEA, TNO, and partially Polito) 3

4 The Correlation study: overall approach technical correlation exercise This contains two sub-steps: Determining CO 2 emissions evaluated on the NEDC and on the WLTP for a range of vehicle configurations and technology packages; Determination of a generalised correlation function between CO 2 on the NEDC and the WLTP; correlation of CO2 emissions measured on the NEDC and on the WLTP on the basis of agreed criteria ensuring comparable stringency using insights from the technical correlation exercise 4

5 The technical correlation exercise Two possible approaches: testing vs. modeling/simulations A mixed approach is adopted: mostly based on modeling combined with results from vehicle testing Pros: cost effective combined pros of both approaches while avoiding many of the cons improves acceptance and robustness of modeling results allows evaluation of aspects of the tests that cannot be adequately simulated allows the extrapolation of test results to different vehicles/technology configurations 5

6 Steps in the technical correlation exercise Definition of a range of vehicle configurations: Starting point is the most recent vehicle registration database available for Europe The car market is divided into different technology-based segments For each segment a representative vehicle model is selected The simulation model for each vehicle is developed Using the model, NEDC- and WLTP-based CO2 emissions are evaluated on different combinations of key vehicle parameters (mass, power, transmission, etc.) and technology options Simulation results are used to estimate the model developed as WLTP/NEDC CO2 correlation function 6

7 Project outline Market segmentation Vehicle/technology selection (SGS) Vehicle testing and modeling Calibration and validation (LAT) Addition of technology packages (LAT) Model analysis and simplification Meta-model development and analysis (JRC/TUG/SGS) NEDC-WLTP Correlation 7

8 Vehicle testing and modelling In order to ensure high simulation quality and to have the possibility to check the results, as many vehicle tests as possible are required. Step 1: Measurement. Vehicles for calibration. Aim: All main technologies from list should be covered (but not all segments needed). Step 2: Simulation task. All technologies and interdependencies. to cover all segments and technologycombinations on the market (current and future). Step 3: Measurement. Validation of simulation results and later of the meta model Additional segments and vehicles types needed, which differ from calibration to check results.

9 Number of vehicles Measurement Programs (M1+N1) UK 30 DE2 DE1 FR 25 NL JRC/OEM SE 10 OEM/TUG 5 LAT (Old+New) 0 Calibration Validation

10 Vehicle list M1 No. Vehicle Measured by Test status Simulated by Model status 1 Opel Astra 1.3D MT LAT Measured LAT Completed 2 Opel Mokka 1.7D AT LAT Measured LAT Completed 3 VW Polo 1.2TSI MT LAT Measured LAT Ongoing 4 Citroen C5 2.0D AT LAT To be measured LAT 6 BMW X1 2.0D MT TUG Measured LAT Completed 7 BMW 116i 1.6G TUG Measured LAT 8 SMART For Two AT TNO Measured LAT Completed 9 AR Mito 1.4 DCT6 FIAT + JRC Measured at FIAT JRC 10 AR Giulietta 1.4 MT6 FIAT + JRC Measured at FIAT + JRC JRC Ongoing 11 Audi A4 MT JRC Measured LAT Ongoing 12 Volvo V40 JRC Measured LAT Ongoing 13 FIAT L MT5 FIAT + JRC Measured at FIAT + JRC JRC Ongoing 14 FIAT L AT5 FIAT + JRC Measured at FIAT JRC 15 FIAT Punto 1.4L MT5 FIAT + JRC Measured at FIAT + JRC JRC Completed 16 Toyota Avensis LAT Measurements available LAT Ongoing 17 Peugeot 308 LAT Measurements available LAT Ongoing 18 Toyota Auris Hybrid LAT (+Toyota) Measurements available LAT Ongoing 19 Toyota Prius PHEV LAT + JRC Measurements available LAT Ongoing 20 Opel Zafira 1.6 MT EU6 Diesel TNO Measured LAT Ongoing 21 VW Crafter DEKRA Measured LAT 22 Audi A8 AT DEKRA Measured LAT 23 BMW 328i AT DEKRA Measured LAT 24 VW Sharan 1.4 TSI DEKRA To be measured LAT 25 Ford Focus 1.0 TÜV Nord Measured LAT 26 Mercedes E350 Lean-Burn TÜV Nord Measured LAT Ongoing 27 PHEV - Porsche Panamera TUV Nord Measured LAT 10

11 META-MODEL DEVELOPMENT 11

12 Development boundaries for Metamodel 1. As simple as possible 2. Minimum number of input variables (and if possible already available in TA procedure) 3. Accuracy of the ΔWLTP-NEDC in the order of 1-2g 4. Minimize statistical approaches and calculations 12

13 Situation by November 2014 (1/2) During the past SGS meetings various ideas and concepts were discussed, most important ones: JRC: Fully physical approach based on key vehicle characteristics, default engine type specific parameters for engine parameterization and optimization based on WLTP test results (WLTP test results used for improving results but are not necessary for running the model) TUG: Fully physical approach based on key vehicle characteristics, WLTP test results, equations linking power to fuel consumption (WLTP test results are necessary for running the model) VW: A semi physical-statistical approach where some lesser vehicle characteristics are necessary but instantaneous fuel flow is used as input (necessary for correlating fuel flow to vehicle speed) Concepts from other OEMs/stakeholders (eg BMW, PSA) were discussed and certain features will be included in the final Metamodel 13

14 Situation by November 2014 (2/2) Initial JRC analysis showed typical ranges for absolute error: 0.5-4g for JRC approach (optimization possibility not considered) 1-6g for TUG approach (with less data were available for improving the model*) 2-4g for VW approach (tested recently only on 2 vehicles so far*) Data from the WLTP (or NEDC if used in opposite direction) shall be used for improving accuracy Use of instantaneous data for Fuel consumption and RPM from the TA test offers the possibility to significantly improve the accuracy. Most input parameters are the same for all 3 options One uniform MM is possible incorporating the best features of all approaches 14

15 JRC approach Overall Results Summary Summary of 33 cold hot combinations tested on 3 vehicles Individual tests performed on other 3 vehicles similar results 15

16 JRC approach Stress test over different simulation cases The model is able to reproduce the CO2-mass relationship An (unexpected) overall match between monitoring data and JRC MM predictions

17 Optimizer concept Example of optimization function in JRC approach (test case gasoline turbo PFI): ΔCO 2 17

18 VW approach preliminary results (JRC internal analysis) First concept investigation, very positive results Significant space for improvement particularly for cold start conditions If RPM signal is used for FC correlation instead of vehicle speed the approach becomes very much similar to JRC proposal Will the instantaneous FC and RPM signals be available? 18

19 VEH 2 VEH 1 TUG approach preliminary results 19

20 COMPAS simulation environment (Developed by JRC) GB/GS model Techn. models Power-RPM model WLTP test-results Fuel consumption model 20

21 COMPAS simulation environment (Developed by JRC) GB/GS model Techn. models Engine map parametrization (JRC) Power-RPM model Test result based Willnas equivalent - VELINE (TUG) Instantaneous FC (VW) WLTP test-results Fuel consumption model 21

22 Key features COMPAS comprises of 2 main calculation modules Power RPM module (more or less the same in all approaches) Simple longitudinal dynamics (WLTP-GTR) Engine power and RPM calc 1hz If RPM not provided from OBD Inclusion of Mech or Elec. loads where needed Generic start-stop logic FC module 2 possibilities for calculation of FC 1. Indicative instantaneous (JRC or VW approach) 2. Mean value based (JRC or TUG based approach) Semi-physical empirical cold start model Calibration - Optimization based on WLTP results Possibility to use instantaneous FC? Accurate calculation of average / instantaneous power wheel; most crucial factor affecting accuracy of all similar models The final selection of exact details will be based on: Best accuracy compared with results obtained from the Cruise simulations by LAT Possibility to use instantaneous FC signal Other possible needs (eg ability to calculate from NEDC 22 to WLTP)

23 Work on COMPAS environment already started First implementation of COMPAS platform in Python code & Excel as GUI Support modules for Cruise and COMPAS test cases running common data generation collection basis Additional tools being tested and further developed: Power-RPM calculation module Optimizer for A/T gearboxes Engine parameterization optimization module Mapping data synchronizing and fitting module in case instantaneous signals become available 23

24 Development needs Time. In order to develop a fully operational tool, JRC/TUG need time to work on the different additional components the meta-model needs to deal with all the technology options. The tool is expected to be fully operationalized by Dec (with updates and checks foreseen up to 2019) Data. The availability of a database of technology-specific maps will make the MM more robust and accurate (although the optimization loop can support in this direction). A first set of 6 maps is being purchased by FEV. More will come. OEMs contribution would be very much appreciated The availability of the FC signal from OBD (as in the case of the VW approach) could be used for the JRC approach as well 24

25 Expected Road-map 2015 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Testing LAT Modeling Analysis Meta-model Finalization of the correlation tool 25

26 NEDC equivalent results WLTP measures CO 2 MPAS metamodel IMPLEMENTING THE WLTP/NEDC BACK- TRANSLATION 26

27 Background During the TWG meeting held in Brussels on May 22, 2014, it was agreed that, at least for the transition period , WLTPbased CO2 figures (measured at type-approval) will be translated in the equivalent NEDC-based ones and then used in the monitoring process in order to keep the CO2 emission targets set by Regulations 443/2009 and 510/2011 (backtranslation approach). W LTP m easures To implement this approach two main elements are required: N EDC equivalent result s A physical model able to translate vehicle-specific WLTP-based CO2 emissions into the corresponding NEDC one (meta-model) A strategy to deal with the operational implications of the backtranslation and with the different family/extension concepts 27 CO 2MPAS m eta- m odel

28 TA WLTP CO 2 determination Given a vehicle family, the test for the determination of the CO 2 is carried out on the vehicle H (representing the worst-case conditions for CO 2 emissions) and, upon request of the OEM, also on the vehicle L (the best-case conditions) The results from the two tests allow the definition of an interpolation line (and the related formula) from which the value of each vehicle put on the market can be determined on the basis of its actual configuration between the L and H At TA, the impact of the different options in affecting the RLs is also certified so that, on the basis of the different options applied to a vehicle, its CO 2 value can be easily determined at the COC W LTP m easures N EDC equivalent result s 28 CO 2MPAS m eta- m odel

29 Test procedure for Vehicle Family Approval using WLTP OEM data Vehicle technical features Unladen Mass(es) Maximum Laden mass(es) Options Mass RLs for Vehicle L RLs for Vehicle H GS tool inputs Options impact on RLs Pre-test calculations Test mass low TM L Test mass high TM H GS pattern Testing Chassis dynamometer set-up for lowest CO 2 case WLTP Type 1 test for lowest CO 2 case Chassis dynamometer set-up for highest CO 2 case WLTP Type 1 test for highest CO 2 case Key output CO 2 values for individual vehicle CoC OEM Provided data Calculations CO 2 H-CO 2 L-CO 2 WLTP Family Approval CO 2 values Vehicle s specific configuration Calculation of position between L and H Application of Vehicle Family WLTP CO 2 values L H {M,F 0,F 1,F 2 } WLTP WLTP Vehicle CO 2 value

30 W LTP m easures Different grouping concepts WLTP family concept According to the WLTP GTR (par ), [ ] unless vehicles are identical with respect to the following vehicle/powertrain/transmission characteristics, they shall not be considered to be part of the same CO 2 family: N EDC equivalent result s CO 2MPAS m eta- m odel Type of internal combustion engine: fuel type, combustion type, engine displacement, full-load characteristics, engine technology, and charging system shall be identical [ ]; Operation strategy of all CO2-influencing components within the powertrain; Transmission type (e.g. manual, automatic, CVT); n/v ratios (engine rotational speed divided by vehicle speed). [ ] the transmission ratios of the most commonly installed transmission type is within 8 per cent; Number of powered axles; 30

31 Different grouping concepts NEDC CO2 extension mechanism According to Annex I of Regulation 692/2008, [ ] The type-approval shall be extended to vehicles differing with regard to the following characteristics, if the CO 2 emissions measured by the technical service do not exceed the type-approval value by more than 4 % for vehicles of category M and 6 % for vehicles of category N: reference mass, technically permissible maximum laden mass, type of bodywork as defined in Section C of Annex II of Directive 2007/46/EC (e.g. sedan, estate, coupé), overall gear ratios, engine equipment and accessories. W LTP m easures N EDC equivalent result s 31 CO 2MPAS m eta- m odel

32 W LTP m easures Implications and strategy Each OEM may adopt different strategies to deal with the two grouping concepts. In order to ensure a level playing field it is necessary to provide a generic approach linking the two grouping concepts. A possible viable solution is to consider the WLTP family concept and the WLTP type-approval process as reference inputs: the way adopted by the OEM to type approve a vehicle family under WLTP will determine the corresponding values for the estimated NEDC CO2 values. The opposite approach would be practically impossible. N EDC equivalent result s 32 CO 2MPAS m eta- m odel

33 Summary The OEM will decide the composition of the WLTP CO 2 family Depending on the composition of the WLTP CO 2 family it may correspond to one or more NEDC type, variant, version approvals including extensions W LTP m easures N EDC equivalent result s 33 CO 2MPAS m eta- m odel

34 Back-translation Two cases are possible 1. The WLTP vehicle family is composed by a vehicle model with no body-shape variations (only variation in CO2 relevant optional and tires) 2. The WLTP vehicle family is composed by a model with different body shape variations (e.g. sedan, stationwagon, cabriolet, coupe, etc.) The determination of the NEDC equivalent vehicle will be different in the two cases W LTP m easures N EDC equivalent result s 34 CO 2MPAS m eta- m odel

35 Case 1 No body-shape variations W LTP m easures WLTP vehicle H represent the worst-case scenario (TMH, worst-case tires, worst aerodynamic configuration) WLTP vehicle L represent the best-case scenario (TML, best-case tires, best aerodynamic configuration) Since vehicle options are not considered in the NEDC, there will be only one equivalent NEDC vehicle N EDC equivalent result s NEDC test mass RR of the worst tires (or second worst depending on the number of tires allowed) AR of the vehicle with no options installed 35 CO 2MPAS m eta- m odel

36 OEM Provided data Calculation of the NEDC-equivalent CO2 value Unladen Mass WLTP RLs for Vehicle L WLTP RLs for Vehicle H RLs-related procedure differences WLTP GS tool inputs Vehicle technical features Pre-test calculations NEDC Test mass NEDC F2 NEDC F0 and F1 WLTP GS pattern Meta-model application WLTP results vehicle L WLTP results vehicle H WLTP/NEDC Meta-model Key output NEDC Family CO 2 value CO 2 values for individual vehicle CoC NEDC vehicle CO 2 value

37 Case 2 Different body-shapes W LTP m easures WLTP vehicle H represent the worst-case scenario (TMH, worst-case tires, worst aerodynamic configuration) for the different body shapes WLTP vehicle L represent the best-case scenario (TML, best-case tires, best aerodynamic configuration) Each body-shape will lead to a different NEDC vehicle An approach like that adopted for WLTP can be considered: the back-translation is only applied to best- and worst-case vehicles and then an interpolation equation (depending on mass and RLs) can be used to any NEDC vehicle laying in between N EDC equivalent result s 37 CO 2MPAS m eta- m odel

38 OEM Provided data Calculation of the NEDC-equivalent CO2 values Minimum Unladen Mass Maximum Unladen Mass WLTP RLs for Vehicle L WLTP RLs for Vehicle H GS tool inputs Impact of options on RLs Vehicle technical features Pre-test calculations NEDC Test mass low IC L NEDC Test mass high IC H NEDC RLs for vehicle L NEDC RLs for vehicle H WLTP GS pattern Meta-model application Key output WLTP results for vehicle L WLTP results for vehicle H NEDC Vehicle L CO 2 WLTP/NEDC Meta-model NEDC Vehicle H CO 2 CO 2 values for individual vehicle CoC OEM Provided data Calculations Vehicle s specific configuration Calculation of position between L and H Application of Vehicle Family WLTP CO 2 values NEDC vehicle CO 2 value

39 TA W LTP m easures All the information required to run the meta-model are already available for the WLTP TA The TA authority could easily run the meta-model to identify the CO 2 emissions corresponding to 1 or 2 NEDC equivalent vehicle(s) If needed, using the same rationale introduced for the WLTP, NEDC CO 2 emissions per each vehicle in the family might be calculated from the interpolation of the values obtained for L and H vehicles This would allow a limited and simple application of the meta-model with a consequent limited impact of the backtranslation to the normal type-approval procedure N EDC equivalent result s 39 CO 2MPAS m eta- m odel

40 Conclusions (1) The back-translation approach is aimed at identifying the NEDC-equivalent of the CO 2 emissions assigned to a vehicle after the type-approval carried out on the WLTP The method requires to achieve the high accuracy in the estimation as well as not to introduce significant burden into the type approval process The proposed approach identifies, per each WLTP-family 1 or 2 equivalent NEDC vehicles. In case more NEDC vehicles can be identified, their CO 2 value is derived from an NEDCbased CO 2 interpolation formula. W LTP m easures N EDC equivalent result s 40 CO 2MPAS m eta- m odel

41 Conclusions (2) The meta-model is only applied to (max) two vehicles over the entire family during the TA process where the necessary inputs are already available The meta-model is applied only when WLTP measurement data are available (important to improve the accuracy) Both the WLTP and the NEDC interpolation formulae are then used with the same logic for the identification of the CO 2 emissions to include in the vehicle certificate of conformity (CoC) W LTP m easures N EDC equivalent result s 41 CO 2MPAS m eta- m odel