Chip Simulation for Virtual ECUs

Transcription

1 Chip Simulation for Virtual ECUs QTronic User Conference 2018 Virtual ECUs and Applications 18th of October, Berlin, Germany Dr. Yutaka Murata Honda R&D Co., Ltd. Automotive R&D Center 1

2 Contents Background Concept of model based simulation environment Engine simulation model - ECU model - Combustion model - Catalyst model RDE simulation combined with vehicle simulation model Summary 2

3 2018 model new Civic Fuel economy (CO 2 ) Modified NEDC 1.6L diesel engine 91 g/km (6MT, Sedan) 93 g/km (6MT, Hachback) 109 g/km (9AT) Exhaust emissions Euro6d-TEMP Civic 3 Passed RDE regulation and achieved 91 g/km

4 RDE definition Chassis dynamometer RDE Vehicle speed profile Fixed Depends on vehicle, driver, route, and traffic Environment (Ta, Pa) Fixed Depends on season, weather, wind, and altitude Road load force Straight, w/o gradient (w/o PEMS) Repeatability with w/o Depends on curves, altitude, road surface, passengers, and baggage (with PEMS) 4 Difficulty to check RDE performance at all conditions during development

5 Method for calibration and validation Validation in real world Tests on road Chassis dynamometer (vehicle) + vehicle simulation Engine test bed (engine) + vehicle simulation EiL (Engine in the Loop) Easy to simulate and calibrate Model (engine) + vehicle simulation MiL (Model in the Loop) PEMS Vehicle simulation: consideration of road load force change due to curves, altitude, and road surface (weather, wind) 5 Necessity of model utilization for efficient development

6 Contents Background Concept of model based simulation environment Engine simulation model - ECU model - Combustion model - Catalyst model RDE simulation combined with vehicle simulation model Summary 6

7 Flowchart of model utilization Vehicle simulation simulation (NEDC, WLTC, RDE) etc. Vehicle model Driver model Route model Traffic model Ne, Te, Gear, V, etc. Experiment definition Boundary finder Dynamic DoE Transient measurement (environmental engine bench) Dynamic statistical combustion model Engine simulation ECU model Catalyst model Synthetic gas flow test bed Calibration target Base maps, environmental corrections, controllers, aftertreatment control etc. Silver software.hex Simulation and optimization Verification Vehicle test.a2l Coupling of vehicle simulation and engine simulation.map 7

8 Engine simulation model Vehicle simulation Ne Te Tw Ta Pa Gear V ECU model Combustion model (Dynamic data based statistical model) Inputs: Ne (Engine speed) Te (Brake torque) Main injection timing Fuel injection pressure VGT opening HP-EGR valve opening LP-EGR valve opening Intake throttle valve opening Intake shutter valve opening Tw (Coolant temperature) Ta (Ambient temperature) Pa (Ambient temperature) Combustion mode signal Outputs: Emissions, temperatures, etc. including sensor values fed back to the ECU model Catalyst model (LNT physical model) Engine outlet emissions Tailpipe emissions 8 Combination of ECU, combustion, and catalyst models

9 Contents Background Concept of model based simulation environment Engine simulation model - ECU model - Combustion model - Catalyst model RDE simulation combined with vehicle simulation model Summary 9

10 Chip simulation for virtualize an ECU Control specification Scheduler.obj.txt HEX for ECU Variables setting MAP file.hex.a2l.map spec. txt.mexw64 MATLAB/Simulink S-function Driver file (CAN, AD, etc.) Real ECU Virtual ECU (Chip simulation) Control logic part (Reuse possible) Driver part (Reuse impossible) 10 Simulation based on HEX without control model and C code

11 Combustion modeling approach Physical model (0D-3D) Statistical model (Empirical model, DoE model) Use case: concept study, advanced research Necessity of parameters tuning based on measurement data Higher number of adjustment parameters High predictive accuracy even at model extrapolation region High dimension -> low calculation speed Suitable for engine hardware development and phenomenological analysis Use case: calibration, validation Necessity of engine hardware and training data Lower number of fitting parameters High predictive accuracy at model interpolation region High calculation speed Suitable for model based engine calibration (optimization) Statistical model is suitable for efficient calibration at later stage of development 11

12 System output System output System input Current time step System input Current time step Advantage of dynamic DoE model Steady state DoE model Dynamic DoE model Measured input Measured input Steady-state model Dynamic model Measured output Model output Measured output Model output Time Time Steady state prediction Model fitting based on averaged measurement data Transient prediction including time lag of measurement apparatus Model fitting based on recorder measurement data Dynamic behavior expression considering histories of system input and output 12

13 Coolant temperature LP-EGR fraction Fresh air mass Boost pressure Fuel injection pressure Main injection timing Fuel injection quantity Main injection timing Engine speed Dynamic DoE for combustion model Upper boundary Lower boundary Space filling test design including steady state test design Time (sec) 13 Space filling methodology for the Gaussian Process Modeling (GPM)

14 Dynamic modeling for engine combustion Model structure for learning time dependent behavior: Regression model with additional inputs and outputs from past time steps Input Output Reference: T. Huber, M. Hanselmann, and T. Kurse: Use of Data Based Models to Predict Any RDE Cycles - Challenges, Experiences and Results, 8th Emission Control Conference, Dresden (2016) Identification of nonlinear and time-dependent system 14

15 gfuel (g/s) Soot (mg/s) NOx (mg/s) Vehicle speed Predictive accuracy of engine model Urban Rural Motorway NOx (mg/s) Measurement Simulation Engine outlet Tailpipe Soot (mg/s) Measurement Simulation Model input: engine speed, brake torque, coolant temp., ambient temp., and ambient pressure gfuel (g/s) Measurement Simulation 15 Achievement of quantitative emissions prediction at RDE

16 Contents Background Concept of model based simulation environment Engine simulation model - ECU model - Combustion model - Catalyst model RDE simulation combined with vehicle simulation model Summary 16

17 自車情報 レーン情報 将来予測 行動指針決定 速度生成 ( 横制御 ) 経路生成 ( 縦制御 ) 軌道 RDE-compliant virtual engine calibration Requirement Regulation change Requirement analysis Vehicle simulation Allocate req. to function ECU modeling Engine emission simulation Virtual engine calibration Verification Powertrain dynamic test bed Validation WLTC RDE H/W Measurement H/W modeling Virtual prototype validation Vehicle 17 Utilization of virtual calibration and validation

18 Vehicle simulation for RDE route Route Rural Digital map Simulation Motorway Urban Altitude Urban Rural Motorway 18 Generation of vehicle speed by vehicle, driver, route, and traffic models

19 Engine speed (rpm) Torque (Nm) Altitude (m) Vehicle speed (km/h) Model based RDE performance evaluation Vehicle, driver, route, and traffic models Vehicle simulation Vehicle speed, engine speed, brake torque, and gear shift position Engine simulation NOx, Soot, CO 2 etc. Altitude Rural Vehicle speed Tailpipe NOx Motorway Gear shift position Engine speed Brake torque Urban Time (sec) 19 Achievement of emission prediction with vehicle and engine simulation

20 Evaluation of emission robustness Simulation Total (Urban+Rural+Motorway) 6MT 9AT Total (Urban+Rural+Motorway) Urban 6MT 20 Validity confirmation of hardware selection and calibration data settings

21 Summary It is a challenge to sufficiently validate RDE performance under all conditions through road tests during vehicle development due to wide range of validating conditions. A model based development technology was established to simulate, verify and calibrate the emissions performance of a vehicle. RDE performance could be accurately predicted by coupling a vehicle driving simulation with an engine simulation that includes an ECU model, combustion model (dynamic data based statistical model), and exhaust aftertreatment catalyst model. Use of the simulation model enabled robust validation of RDE performance under various conditions that assume driving on actual roads. 21 Thank you very much for your kind attention.

22 ** 22 **