TTX 2013

Slide 1
Estimation of Road Load Parameters via On-road

Vehicle Testing
Dr. Rahul Ahlawata, Dr. Jürgen Bredenbeckb & Mr. Tatsuo Ichigec
a A&D Technology, Michigan, USA b A&D Europe GmbH, Griesheim, Germany
c A&D Company, Tokyo, Japan
Tire Technology Expo 2013
February 5-7, Cologne, Germany
Proprietary and Confidential, A&D Technology, Inc. www.aanddtech.com

Slide 2
Energy Loss in Vehicles

Standby Accessories
17%(4%) 2%(2%)
Aerodynamic Drag
3%(11%)
Fuel Tank Rolling Resistance

100%(100%) 4%(7%)
Braking
6%(2%)
Engine Loss Drivetrain Loss

62%(69%) 6%(5%)
Losses of fuel energy in a vehicle in city usage (highway usage)

[U.S. National Academy of Science, 2006]

Slide 3
Energy Loss in Vehicles

Standby Accessories
17%(4%) 2%(2%)
Aerodynamic Drag
3%(11%)
Fuel Tank Rolling Resistance

100%(100%) 4%(7%)
Braking
6%(2%)
Engine Loss Drivetrain Loss

Vehicle
62%(69%)
Road Load
6%(5%)
Road Grade
Road load is define as the “…force or torque which opposes the movement of
a vehicle…”
[ISO 10521-1, 2006]
Slide 4
Significance of Road Load Determination

Energy
Vehicle platooning [1,2] Roll-over prevention [3] management [4,5]
[SARTRE project test, Sweden]
Offline modeling & [Toyota]

control development [6]
Engine certification [7]
[Mathworks]

Slide 5
Objective
Estimate the road load parameters of a vehicle

 Focus on rolling resistance & aerodynamic drag
 Use on-road testing of a production vehicle
 Use a novel force measurement method & compare results
with traditional method(s)

Slide 6
Outline
1. Introduction
2. Road load fundamentals
3. Instrumentation &
sample data
4. Coast down method
5. Force measurement
method
6. Summary
Slide 7
Rolling Resistance
For a free rolling tire under no slip:
Mg: Vertical load on the tire due to sprung & unsprung mass
vx: Tire longitudinal velocity ω: Tire angular speed Rz: Ground reaction force
Rx: Rolling resistance force Mrr: Rolling resistance moment r: Loaded tire radius
vx vx vx
Mg ω Mg ω Mg ω
Mrr
Rx Rx
Tire normal force a
distribution is asymmetric.
The resultant normal force Rz Rz
acts towards the leading a·Rz = Mrr= r· Rx Rolling resistance
edge.
Proprietary and Confidential, A&D Technology, Inc. www.aanddtech.com force
Slide 8
Aerodynamic Drag
[BMW]
𝐹𝑑𝑟𝑎𝑔 = 0.5 ∙ 𝜌 ∙ A ∙ 𝐶𝑑 ∙ (𝑣𝑟𝑒𝑙 )2
Density of air
Frontal area of the vehicle
Coefficient of aerodynamic drag
Relative velocity of the vehicle wrt the wind
Slide 9
Total Road Load

The most commonly used form of road load equation is:
𝐹𝑟𝑜𝑎𝑑 𝑙𝑜𝑎𝑑 = 𝑎 + 𝑏 ∙ 𝑣𝑥 + 𝑐 ∙ (𝑣𝑟𝑒𝑙 )2 + 𝑀 ∙ 𝑔 ∙ sin(𝜃)
Predominantly Includes Includes influence

includes the aerodynamic drag of road grade
effect of rolling
resistance Includes dependence
of rolling resistance on
velocity & drivetrain
losses
• ‘b’ term is not always included
• A number of other formulations exist [8], including
• Influence of rotational inertias
• Correction factors
• Additional dependencies
Slide 10
Road Load Measurement Methods

1. Individual component measurements
[Terrametrix]
[Nissan]
[Volvo]
[Maptek]
Rolling resistance: Tire Aerodynamic drag:
testing Wind tunnel testing Grade: Road profiling
Pros: High repeatability, aids in parametric evaluation

Cons:
Proprietary Higher
and Confidential, cost, Inc.
A&D Technology, may not represent real driving conditions
www.aanddtech.com
Slide 11

2. Coast-down method [8,10] 3. Torque measurement [9]
[VTI Sweden] [Timken]
[9]
Pros: Less instrumentation Pros: Drivetrain losses are excluded

Cons: Time consuming tests, include Cons: Difficult to install on
drivetrain losses production vehicle
Slide 12

4. Complete vehicle measurement system
[A&D]
[A&D]

Slide 13
On-road Vehicle Testing
Stop
Test Vehicle: FWD Mini Cooper S Start
Test Track: Proving ground in Tochigi, Japan

Slide 14
Instrumentation
Anemometer
Measures wind velocity & direction
6-Component Wheel Force Transducer
(Fx, Fy, Fz, Mx, My, Mz)
o Distributed force bridges with model based decomposition to get
orthogonal force components
o Very low cross sensitivity, interference & temperature sensitivity and
high sampling rate
oA&D0.1%
Proprietary and Confidential, resolution
Technology, Inc. (6N or 1.8Nm) www.aanddtech.com
Slide 15
Instrumentation
Anemometer
Measures wind velocity & direction
6-Component Wheel Force Transducer
(Fx, Fy, Fz, Mx, My, Mz)
Influence of drivetrain losses is included in this measurement

Influence of a and b terms is included in this measurement (ignoring bearing
friction and wheel well aerodynamic losses) www.aanddtech.com
Proprietary and Confidential, A&D Technology, Inc.
Slide 16
Vehicle Instrumentation II
Laser Doppler Sensor
Measures vehicle velocity & slip
angle
GPS Sensor & In-vehicle Network
Measures vehicle longitude,
latitude, altitude, and ECU CAN
communication
Inertial Sensor
Measures vehicle roll,
pitch and yaw
Wheel Position Sensor

Measures 6 degrees of freedom of the tire
Digital Signal Processing

& Acquisition
100Hz sampling
(max 100kHz)
Slide 17
Sample Data
FL
FR
RL
RR

Slide 18
Sample Data
Acceleration Coast-down Acceleration Braking
FL
FR
RL
RR

Slide 19
Sample Data
Acceleration Coast-down Acceleration Braking
Contribution of a and b terms
FL
FR
RL
RR

Slide 20
Outline
1. Introduction
sample data
method
6. Summary
Slide 21
Coast-down Tests
Procedure:
• Conduct tests on a flat road with low wind conditions
• Accelerate the vehicle and put the transmission in N
• Begin coast down in a straight line
• Record vehicle velocity and vehicle velocity relative to wind as a function
of time
35
30
25
20
Velocity (m/s)
Coast-down
15
10
-5
0 10 20 30 40 50 60 70 80 90
Proprietary and Confidential, A&D Technology, Inc. Time (sec) www.aanddtech.com
Slide 22
Coast-down Tests
Procedure:
of time

Flat road assumption
6
Road grade (deg)
4
2
Road grade (deg)
-2
-4
-6
20 Technology,
Proprietary and Confidential, A&D 40 60
Inc. 80 100 120 140 160 180
www.aanddtech.com 200
Time (sec)
Slide 23
Coast-down Tests
Procedure:
of time

𝑑𝑣𝑥
−𝐹𝑟𝑜𝑎𝑑 𝑙𝑜𝑎𝑑 = 𝑀
𝑑𝑡
• Use linear regression to obtain coefficients a, b & c
• Use minimization of ∥ 𝐿 ∥2
• Verify by Simulated Annealing
• SAE J1263, J2263 and ISO 10521-1 contain more detailed procedures

Slide 24
Coast-down Test Results

1200
Measured road load force
Regression estimates
1000
Coefficient of • Front tires: Bridgestone
800
determination, R2=0.9087 Sneaker
600 • Rear tires: Bridgestone
Force (N)
400 Sneaker
200
• Estimated Values:
Data-sets augmented a = 194.87, b = 3.87, c = 0.37
0
-200
0 10 20 30 40 50 60 70 80 90
Time (sec)
35
(33m/s, 118km/hr, 73mph) Measured velocity
95% Confidence bounds: 30

Calculated velocity
a: 184 – 204 25
Velocity (m/s)
20
b: 2.7 – 5
15
c: 0.35 – 0.39
10
5
(7.5m/s, 27km/hr, 16mph)
Proprietary and Confidential, A&D Technology, Inc. 0 10 20 www.aanddtech.com
30 40 50 60 70 80 90
Time (sec)
Slide 25
Coast-down Test Results

1200
Measured road load force
Regression estimates
1000
• Front tires: Bridgestone
Coefficient of
800
2 Blizzak
determination, R =0.9280 • Rear tires: Bridgestone
600
Force (N)
400
Sneaker
• Estimated Values:
200
Data-sets augmented a = 191.69, b = 2.54, c = 0.41
0
-200
0 10 20 30 40 50 60 70 80
Time (sec)
35
Measured velocity
95% Confidence bounds: 30

Calculated velocity
a: 183 – 200 25
Velocity (m/s)
20
b: 1.5 – 3.5
15
c: 0.39 – 0.43
10
5
Proprietary and Confidential, A&D Technology, Inc. 0 10 20 30 40
www.aanddtech.com 50 60 70 80
Time (sec)
Slide 26
Validation Procedure
𝐹𝑟𝑜𝑎𝑑 𝑙𝑜𝑎𝑑 = 𝑎 + 𝑏 ∙ 𝑣𝑥 + 𝑐 ∙ (𝑣𝑟𝑒𝑙 )2
𝑑𝑣𝑥
During coast-down: −𝐹𝑟𝑜𝑎𝑑 𝑙𝑜𝑎𝑑 = 𝑀
𝑑𝑡
Generalized equation under all conditions (including coast-down):
𝑑𝑣𝑥
Tire traction/braking
Σ𝐹𝑡𝑏 − 𝐹𝑟𝑜𝑎𝑑 𝑙𝑜𝑎𝑑 = 𝑀
force 𝑑𝑡
2 𝑑𝑣𝑥
Σ𝐹𝑡𝑏 − (𝑎 + 𝑏 ∙ 𝑣𝑥 + 𝑐 ∙ 𝑣𝑟𝑒𝑙 ) = 𝑀
𝑑𝑡
2 𝑑𝑣𝑥
(Σ𝐹𝑡𝑏 −𝑎 − 𝑏 ∙ 𝑣𝑥 ) − 𝑐 ∙ 𝑣𝑟𝑒𝑙 = 𝑀
𝑑𝑡
2 𝑑𝑣𝑥
Wheel force Σ𝐹𝑥 = 𝑐 ∙ 𝑣𝑟𝑒𝑙 + 𝑀
sensor 𝑑𝑡
measurements
Slide 27
Validation Procedure
35
30
25
20
Velocity (m/s)
15
Parameter
10
5
identification
0
-5
Parameter Validation
0 10 20 30 40 50 60 70 80 90
Time (sec)
Estimates are poor outside the coast-down region

Slide 28
Residual Analysis
2 𝑑𝑣𝑥
Σ𝐹𝑥 = 𝑐 ∙ 𝑣𝑟𝑒𝑙 + 𝑀
𝑑𝑡
2
𝑑𝑣𝑥
𝑅𝑒𝑠 = Σ𝐹𝑥 − [𝑐 ∙ 𝑣𝑟𝑒𝑙 + 𝑀 ]
𝑑𝑡
Analyze cross-correlation coefficient of residuals:
An example:
x1 ∈ 𝒩 0,1
0.1
x2 ∈ 𝒩(0,1) x1 ∈ 𝒩 0,1 x2 = 𝑥13
1.2
Normalized cross correlation coefficient
Normalized cross correlation coefficient

1
0.05
0.8
0.6
0
0.4
-0.05 0.2
-0.1 -0.2
-1000 -800 -600 -400 -200 0 200 400 600 800 1000 -1000 -800 -600 -400 -200 0 200 400 600 800 1000
Time Time

Slide 29
Residual Analysis
2 𝑑𝑣𝑥
Σ𝐹𝑥 = 𝑐 ∙ 𝑣𝑟𝑒𝑙 + 𝑀
𝑑𝑡
2
𝑑𝑣𝑥
𝑅𝑒𝑠 = Σ𝐹𝑥 − [𝑐 ∙ 𝑣𝑟𝑒𝑙 + 𝑀 ]
𝑑𝑡
Analyze cross-correlation coefficient of residuals:
0.6 0.5
0.5
Residuals & steering angle Residuals & accelerator
Normalzed cross-correlation coefficient
Normalized cross-correlation coefficient

0.4
(Residuals & Acc pedal position)

(Residuals & Steering angle)
0.4
0.3
pedal position
0.3
0.2
0.2
0.1
0.1
0
-0.1 0
0 10 20 30 40 50 60 0 10 20 30 40 50 60
Time (sec) Time (sec)

Slide 30
Limitations of Coast-down Tests

Procedure:
1. A long straight flat track is needed
• SAE procedures requires a minimum speed band of 70 to 15 mph
2. Results include drivetrain losses and may not be suitable for some
applications
3. Results are not consistent for all driving conditions, especially outside the
coast-down region
• Residuals are correlated with accelerator pedal position, steering
angle,…
4. Predictor basis is not orthogonal giving rise to the mathematical
complications due to multicollinearity
• Estimates of a, b, c might be biased or have high variance

Slide 31
Outline
1. Introduction
sample data
method
6. Summary
Slide 32
Identification using Force Method

𝑑𝑣𝑥 2
Σ𝐹𝑥 − 𝑀 = 𝑐 ∙ 𝑣𝑟𝑒𝑙
𝑑𝑡
Use regression to identify c c = 0.5371
35 ID ID
30
25
20
Velocity (m/s)
15
10
0
Validation
-5
0 50 100 150 200 250
Time (sec)

Slide 33
Validation Test
R2 = 0.9926
95% confidence limit for
Proprietary and Confidential, A&D Technology, Inc. c: 0.53- 0.55
www.aanddtech.com
Slide 34
Comparison of Coast-down & Force Method
• Front tires: Bridgestone Blizzak

• Confidential,
Proprietary and Rear tires: Bridgestone
A&D Technology, Inc. Sneaker www.aanddtech.com
Slide 35

Slide 36

Slide 37
Cross-correlation of Residuals
0.5
New residuals
0.4
Residuals & steering angle Coastdown residuals
Normalzed cross-correlation coefficient

(Residuals & Steering angle)
0.3
0.2
0.1
-0.1
0.6 -0.2
0.5
Residuals & accelerator
0 50 New residuals
100
Coastdown Time (sec)
residuals
150 200 250
Normalized cross-correlation coefficient
pedal position
(Residuals & Acc pedal position)
0.4
0.3
0.2
0.1
-0.1
-0.2
0 and Confidential,
Proprietary 50A&D Technology,100
Inc. 150 200 250
www.aanddtech.com
Time (sec)
Slide 38
Traction & Braking Scenarios

Assumptions: m: Mass of tire wheel assembly
• Small inclination and side-slip angles J: Polar moment of inertia of tire-wheel
vx • No vertical displacement of the tire
Mg assembly about the center of wheel hub
• No slip T: Applied torque at wheel hub
ω Fa: Tire-road friction force
Z Mrr
T
X
Let WFS measurements be represented as Fx, Fz and My

r Then, From laser Doppler sensor
Fa From encoder
Rz
Rz
Fa
Proprietary and Confidential, A&D Technology, Inc. Determined before the experiment
www.aanddtech.com
Slide 39
Calculation of ‘a’ from Force Method

6 constant speed tests for each speed, 18 tests total
RL Tire
Assuming that wheel well aerodynamic losses are negligible at low

speeds, calculate
𝑎 = Σ𝑅𝑥 a = 271
Note: Measured value of rolling resistance is much higher than what

standardized tests predict [11]
Slide 40
Calculation of ‘b’ from Force Method
Mean RRF=85.4381 N
σRRF= 1.7629 N Mean RRF=84.5861 N
σRRF= 1.9318 N
Mean RRF=84.8949 N
σRRF= 2.2992 N
Very slight reduction in rolling resistance as speed increases
𝛿(Σ𝑅𝑥 )
𝑏 = b≈0
𝛿(Σ𝑣𝑥 )
Slide 41
Summary
Coast-down Method Force Method
a = 191.69, b = 2.54, c = 0.41 a = 271, b = 0, c = 0.5371
• R2 = 0.9280 • R2 = 0.9926
• Estimate variance is higher • Estimation based on physics;
• Estimation only over coast-down; variance is very low
road load is under-estimated for • Estimation can be carried out
non-coast-down conditions during all conditions
• Residuals are correlated with • Correlation is significantly
driver inputs reduced
• Changing the tire changes ‘c’ • Changing the tires preserves ‘c’
substantially very closely
• Influence of drivetrain losses is • Influence of drivetrain losses is
included NOT included
• Less instrumentation is needed • More instrumentation required
• Less distortion of vehicle • Vehicle aerodynamics are
aerodynamic modified to a greater extent
Slide 42
Acknowledgements
• On-road data acquisition team:

– Takayasu Sasaki
– Yuuki Sakurai
– Masaaki Banno
– Hiroki Yamaguchi
• Kenji Sato, A&D Technology, Ann Arbor
• Dr. Michael Smith, A&D Technology, Ann Arbor

Slide 43
References
1. D.Yanakiev & I.Kanellakopoulos, “Speed Tracking and Vehicle Follower Control Design for
Heavy-Duty Vehicles”, Vehicle System Dynamics, Vol. 25, No. 4, 1996
2. D.Yanakiev & I.Kanellakopoulos, “Nonlinear spacing policies for automated heavy-duty
vehicles”, IEEE Transactions on Vehicular Technology, Vol. 47, No. 4
3. Bae, Ryu & Gerdes, “Road grade vehicle parameter estimation for longitudinal control using
GPS”, 2001 IEEE Intelligent Transportation Systems
4. C. Musardo, G. Rizzoni & B.P. Staccia, “A-ECMS: An Adaptive Algorithm for Hybrid Electric
Vehicle Energy Management” 2005 European Control Conference
5. Hong Yang & Joel Maguire, “Predictive Energy Management Control Scheme for a Hybrid
Powertrain System”, US Patent 2011/0066308 A1
6. www.Carsim.com
7. J Fredriksson, E Gelso, M Åsbogard, M Hygrell, O Sponton, NG Vagstedt, “On emission
certification of heavy-duty hybrid electric vehicles using hardware-in-the-loop simulation”,
Chalmers University of Technology, 2011
8. Karlsson, Hammarström, Sörensen & Eriksson, “Road surface influence on rolling resistance
- Coastdown measurements for a car and an HGV”, VTI, 2011
9. J.Żebrowski, “Traction efficiency of a wheeled tractor in construction operations”, Automation
in Construction, Vol. 19, No. 2, 2010
10. Sandburg & Ejsmont, “Noise emission, friction and rolling resistance of car tires – Summary
of an experimental study”, National conference on noise control engineering, Dec 3-5, 2000,
Newport beach, California
11. S.K. Clark, “A handbook for the rolling resistance of pneumatic tires”, 1979
Slide 44
More Information
Email: Visit our booth:

TTX 2013

Caricato da

Informazioni sul documento

Titolo originale

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

TTX 2013

Caricato da

Copyright:

Formati disponibili

Slide 1

Estimation of Road Load Parameters via On-road

Proprietary and Confidential, A&D Technology, Inc. www.aanddtech.com

Energy Loss in Vehicles

Fuel Tank Rolling Resistance

Engine Loss Drivetrain Loss

Losses of fuel energy in a vehicle in city usage (highway usage)

Proprietary and Confidential, A&D Technology, Inc. www.aanddtech.com

Energy Loss in Vehicles

Fuel Tank Rolling Resistance

Engine Loss Drivetrain Loss

Significance of Road Load Determination

[SARTRE project test, Sweden]

Offline modeling & [Toyota]

Proprietary and Confidential, A&D Technology, Inc. www.aanddtech.com

Estimate the road load parameters of a vehicle

Proprietary and Confidential, A&D Technology, Inc. www.aanddtech.com

2. Road load fundamentals

4. Coast down method

𝐹𝑑𝑟𝑎𝑔 = 0.5 ∙ 𝜌 ∙ A ∙ 𝐶𝑑 ∙ (𝑣𝑟𝑒𝑙 )2

Total Road Load

Predominantly Includes Includes influence

Road Load Measurement Methods

Pros: High repeatability, aids in parametric evaluation

Road Load Measurement Methods

[VTI Sweden] [Timken]

Pros: Less instrumentation Pros: Drivetrain losses are excluded

Road Load Measurement Methods

Proprietary and Confidential, A&D Technology, Inc. www.aanddtech.com

On-road Vehicle Testing

Test Track: Proving ground in Tochigi, Japan

Influence of drivetrain losses is included in this measurement

Wheel Position Sensor

Digital Signal Processing

Proprietary and Confidential, A&D Technology, Inc. www.aanddtech.com

Proprietary and Confidential, A&D Technology, Inc. www.aanddtech.com

Contribution of a and b terms

Proprietary and Confidential, A&D Technology, Inc. www.aanddtech.com

2. Road load fundamentals

4. Coast down method

𝐹𝑟𝑜𝑎𝑑 𝑙𝑜𝑎𝑑 = 𝑎 + 𝑏 ∙ 𝑣𝑥 + 𝑐 ∙ (𝑣𝑟𝑒𝑙 )2 + 𝑀 ∙ 𝑔 ∙ sin(𝜃)

𝐹𝑟𝑜𝑎𝑑 𝑙𝑜𝑎𝑑 = 𝑎 + 𝑏 ∙ 𝑣𝑥 + 𝑐 ∙ (𝑣𝑟𝑒𝑙 )2 + 𝑀 ∙ 𝑔 ∙ sin(𝜃)

Proprietary and Confidential, A&D Technology, Inc. www.aanddtech.com

Coast-down Test Results

95% Confidence bounds: 30

Coast-down Test Results

95% Confidence bounds: 30

Estimates are poor outside the coast-down region

Normalized cross correlation coefficient

Proprietary and Confidential, A&D Technology, Inc. www.aanddtech.com

Normalized cross-correlation coefficient

(Residuals & Acc pedal position)

Proprietary and Confidential, A&D Technology, Inc. www.aanddtech.com

Limitations of Coast-down Tests

Proprietary and Confidential, A&D Technology, Inc. www.aanddtech.com

2. Road load fundamentals

4. Coast down method

Identification using Force Method

Proprietary and Confidential, A&D Technology, Inc. www.aanddtech.com

Comparison of Coast-down & Force Method

• Front tires: Bridgestone Blizzak

Comparison of Coast-down & Force Method

Proprietary and Confidential, A&D Technology, Inc. www.aanddtech.com