© The Applied Control Technology Consortium

Hardware in the loop for

Electronic Throttle System
Identification and Control
Dr. Arek Dutka
ACTC
© The Applied Control Technology Consortium

Electronic Throttle Body

Connector pins
-Throttle controls air flow to car engine
Return spring
Ø Controls fuel flow to car engine
Throttle valve
- In conventional cars: pedal linked directly to throttle
Coil link is decoupeled
- With an electronical throttle this mechanical
Pedal sensor

Pedal and Throttle positions decoupling results in:

- Lower fuel consumption
- Better engine torque control
- Different engine response available (normal, snow, mountains)

1
© The Applied Control Technology Consortium
System Overview

Laptop PC with LabVIEW

- VI-Programming & Deploy
- Identification & Control Design
- Remote Control & Monitoring

PWM Control-Signal
H-Bridge
PXI with LabVIEW RT
PWM Driving-Signal
- Signal Generation
Measurements
- Data Acquisition
- Control Algorithm Electronic Throttle
© The Applied Control Technology Consortium

Hardware Connections

2
 © The Applied Control Technology Consortium © The Applied Control Technology Consortium

Steady State Characteristic

Driver (Power Amplifier)

3
© The Applied Control Technology Consortium
System model structure
n The following Hammerstein model structure will be used
for modelling

l Results will be compared with the pure linear model

n The above model structure suggests that the steady-state

characteristic is a part of the system located before the
linear dynamics
© The Applied Control Technology Consortium

System model structure

n The assumption may not be in line with reality
l Steady-state response and dynamics may not be separable
l Dynamics may be non-linear

n Let’s assume that our assumptions are sufficiently

accurate
l Remove the steady-state behaviour by using a pseudo-inverse of
the steady-state characteristic

SS characteristic inverse

4
© The Applied Control Technology Consortium
System model structure
n Assuming that the static non-linearity is cancelled out by
the pseudo-inverse the remaining model of the system is
linear:

SS characteristic inverse
© The Applied Control Technology Consortium

Identification Experiment

n Inverse of the input characteristic scales the input to the

‘virtual linear system’ with the similar input-output range
l This range is about 0.63 … 4.53 [V]

n The random signal is generated and passed through the

inverse of the steady-state characteristic
l Signal that is obtained on the SS Characteristic output is a Duty
Cycle signal that is sent to the power amplifier and the Electronic
Throttle’s coil

5
© The Applied Control Technology Consortium
Identification Experiment

SS characteristic inverse

Input for Hammerstein model identification Input for Linear model identification
© The Applied Control Technology Consortium

Identification Experiment
Hammerstein model identification: Identification Data

6
© The Applied Control Technology Consortium
Identification Experiment
Hammerstein model identification: Validation Data

Mean absolute error: 1.675E-1

Mean squared error: 4.588E-2
© The Applied Control Technology Consortium

Identification Experiment
Linear model identification: Identification Data

7
© The Applied Control Technology Consortium
Identification Experiment
Linear model identification: Validation Data

Mean absolute error: 2.045E-1

Mean squared error: 6.847E-2
© The Applied Control Technology Consortium

More data…

8
 © The Applied Control Technology Consortium © The Applied Control Technology Consortium
More data…

More data…

9
 © The Applied Control Technology Consortium © The Applied Control Technology Consortium

Duty cycle
More data…

Controller: Hand-tuned - PID

*10ms

*10ms

10
© The Applied Control Technology Consortium
Controller: PID from IMC tuning
based on linear model

*10ms

*10ms

Controller: Hammerstein – PI from

© The Applied Control Technology Consortium

Z-N tuning

*10ms

*10ms
SP u0 u y
PI ET

11
© The Applied Control Technology Consortium
Physical Model
Coil resistance changes
with Temperature

Voltage
Throttle Angle

- Neglected electric dynamics -Friction torques ( stiction/Coulomb friction)

- Back electromotive force
- Spring torque

R.Scattoliniet al.: „Modeling and Identification of an electromechanical Internal Combustion

Engine Throttle Body. Control Eng. Practice, Vol. 5, No.9, pp. 1253-1259, 1997.

Physical Model Identification

© The Applied Control Technology Consortium

Result

*10ms

Mean squared error of scaled data:

0.057
• Note: Hammerstein model was 0.026

12
© The Applied Control Technology Consortium
Problems with EKF identification

• The discretisation of the system with discontinuities causes numerical

problems (e.g. Coulomb friction – substituted by tanh() )

• The model contains 6 unknown parameters – simultaneous identification

depends on initial conditions

• Overall, poor knowledge of the initial system parameters which caused

problems

• Unmodelled nonlinearities, observable in the steady state characteristic,

might affect these results
© The Applied Control Technology Consortium

Simulation model

13
© The Applied Control Technology Consortium
Alternative Control Technique
n PID control performance was not good
enough: poor speed of response
lThis was the case for PID controller designed
based on:
nManual tuning
nLinear model based tuning
nNon-Linear model based tuning
n The best result was obtained for
Hammerstein model
lBut with significant amount of control activity
© The Applied Control Technology Consortium

Alternative Control Technique

n Since the throttle is controlled through

PWM modulation, the control signal is
discontinuous and fast-switching
n This suggests that Sliding Mode control
technique might be considered
n We’ll use system simulation to investigate
properties of that control technique

14
© The Applied Control Technology Consortium
Sliding Mode Control
a
Dc = 0 or 1
Dc = 1

a
Dc = 0

Dc = 1
© The Applied Control Technology Consortium

Sliding Mode Control

n The approach used here is by far simpler than the
formal Sliding Mode Control technique
l Based on relay control
l Depending on the current being on/off the throttle angle
rate will increase or decrease
n The sliding manifold will need to be adjusted for
these rates to give stable response
n The controller will be very simple with only 1
tuning parameter

15
© The Applied Control Technology Consortium
Sliding Mode Control
n We’ll use the model identified with the EKF
α& = aα + b b = −aα ref
a Negative number Positive number
1
α s = −αγ
& + αref γ =−
a

a ref a
α − α s = αγ
& − α ref + α < 0
Need to increase speed (increase control)
© The Applied Control Technology Consortium

Sliding Mode Control

n The controller is implemented below:

n The relay detects if control needs to be set

to max. value or minimum value (0)
lIt is also possible to introduce a hysteresis to
reduce switching frequency

16
 © The Applied Control Technology Consortium © The Applied Control Technology Consortium

System simulation: front panel

System simulation: diagram

17
 © The Applied Control Technology Consortium © The Applied Control Technology Consortium

Gamma = 0.5
Hardware results

Hardware results

18
 © The Applied Control Technology Consortium © The Applied Control Technology Consortium

Gamma = 0.05

Gamma = 0.01
Hardware results
Hardware results

19
 © The Applied Control Technology Consortium © The Applied Control Technology Consortium

Gamma = 0.005
Gamma = 0.008

Hardware results
Hardware results

20
 © The Applied Control Technology Consortium © The Applied Control Technology Consortium

Gamma = 0.0005

Thank you!
Hardware results

21