Professional Documents
Culture Documents
Integrated_Plant_and_Controller_Design_of_a_Combin
Integrated_Plant_and_Controller_Design_of_a_Combin
Integrated_Plant_and_Controller_Design_of_a_Combin
DSCC2011
October 31 - November 2, 2011, Arlington, VA, USA
DSCC2011-5959
ABSTRACT decades ago, when the requirements from automation were not
This article proposes a plant and controller design method considered. Therefore, when automatic technologies are applied
based on performance analysis of a header height control to improve the efficiency and accuracy on such legacy designs,
problem on a combine harvester. The achievable bandwidth the plant dynamics can place obstacles to achieving a desired
was found to be limited by the under-actuated and non- performance. The header height control discussed in this paper
collocated features of the mechanical structure, and a is an example of such a problem.
parameter optimization is then applied to solve this problem by
improving the property of open loop zeros and poles. H∞
controller design is used to achieve the best performance with Combine body
NOMENCLATURE
mcom combine body mass. FIGURE 1 COMBINE SYSTEM
mh header mass.
The header height control problem is a long standing issue
dw wheel distance between front and rear tires. in the combine harvester industry. It has been estimated that
approximately 75% of the crop loss occurs at the header [2],
t damping ratio of front and rear tires.
and a significant portion of the header loss is caused by
k spr _ t ratio of tire spring constant to the initial value improper setting of the header height. Figure 1 shows a
schematic of a combine harvester system operating in the
h header height. vertical plane. The header height is defined as the distance
lc hydraulic cylinder length between the header tip and ground. If the header height is too
large, there is a reduction in harvest yield since much of the
I. INTRODUCTION grains will be left un-harvested. Conversely, if the header is
The closed loop system performance of mechanical positioned too low, equipment damage or operator fatigue will
systems is influenced by the plant dynamics which are result due to the impact with the ground. By raising or lowering
determined by the topology of the mechanical structure, the the header with an actuator, usually hydraulic, the header height
parameters of the components, the placement of sensors and can be adjusted.
actuators, the choice of the materials, and other design features
[1]. However, many agriculture machines were designed
has been used in multiple fields such as vehicle control [9] and
aircraft [10], so it is adopted in this paper for the robust where n is the number of the total DOF, m is the number of
controller design. passive DOF, zi , zi are zeros, p j , p j are the poles corresponding
The rest of this paper is organized as follows. Section II to the natural frequency and damping ratio of the n DOF.
briefly introduces the model of the system and discusses the After linearizing the equations at a certain equilibrium point,
limitation caused by mechanical structure. Section III presents the open-loop transfer function from cylinder velocity input to
the parameter design process. Section IV introduces the output height is given by Eq. (4) [5]. It has an order of 5 instead
controller design and the comparison of the closed loop of 6 by Eq. (3). The reason is that in Eq. (3) we assumed the
performance between the original plant and the plant obtained input to the plant is a force corresponding to a pressure control
in section III. Section V gives the conclusions. valve in the electro-hydraulic system; however, in the
experimental combine system, the hydraulic actuator outputs a
II. MODELING AND LIMITATION ANALYSIS velocity by flow control valves. Since there were two
integrators in the denominator with a force input [4], one of
The combine system is simplified as a planar multi-body them is cancelled by the derivative relationship between a
system shown in Figure 2, which contains two rigid bodies: the force input and a velocity input, which results a 5th order
combine body and the header. The tires connecting the body to system in this specific case.
the ground are modeled as spring-damper systems, and the only 2
k s zi s zi
actuator is installed between the header and combine body. This h (t ) (4)
Pcom s L 2
i=1
combine body) and two passive ones (the combine body Using the data from a test combine in [5], the zeros and poles
rotation and vertical translation relative to the center of locations can be obtained in Figure 3. The transfer function
r + Kc y
(5) , and K c is the controller gain. - C ( s) Pcom ( s )
e
m arg P igc
0 Kc (6)
P igc
-5
where is a coefficient close to 1, m is the desired phase
-10 margin, gc is the gain crossover frequency, and P is the plant
transfer function.
-15
-2 -1.5 -1 -0.5 0 0.5
50
Real axis
required controller gain
30
450
20
400
cost function J1
10
350
0
-1 0 1 2
10 10 10 10
300
Achievable crossover frequency (rad/s)
FIGURE 9 ACHIEVALBE BANDWIDTH ANALYSIS FOR
250
DESIGNED SYSTEM
200
0 10 20 30 40 50 60 70 IV. CONTROL DESIGN AND PERFORAMANCE
iteration COMPARISON
FIGURE 7 CONVERGENCE OF J1 The controller used in previous section is chosen by simple
analysis; therefore, a robust controller will be designed in this
TABLE 3. ZEROS AND POLES PROPERTIES FOR section to further improve the closed loop performance. The
DESIGNED PRODUCT *
plant is fixed to be the optimal transfer function Pcom s in Eq.
Undamped natural Damping (9), while the controller Gc is to be designed (Figure 10).
Zeros/Poles
frequency ratio
Fixed plant
z1 , z1 3.6 0.47 r + y
Gc (s)
*
Pcom s
-
z2 , z2 1.2 0.52
p1 , p1 1.63 0.8
FIGURE 10 PARAMETER DESIGN FOR PLANT
p2 , p2 1 0.37
The model uncertainty can be estimated by experiments
and system identifications under different operating conditions.
In this paper, we assume a normalized model uncertainty with
its infinite norm smaller than 1. Then an H controller can be
15 original zeros
original poles used to balance the tradeoff among three frequency domain
zeros after design criteria: robustness, performance and control effort [15]. These
10 poles after design
three design goals can be achieved by a mixed sensitivity
method by evaluating the three key terms in Eq. (10): weighted
5
complementary sensitivity function WI T , weighted sensitivity
Imaginary axis
0
function WP S and control effort constraint Wu Gc S .
WP S
-5
N Wu Gc S (10)
-10 WI T
0.04s 1 still satisfy the design requirements, but Gc0 S0* slightly violates
WI (11)
0.01s 0.95 the constraint of 20dB. The magnitude plot of T0* has a sharp
s 10 drop at 1.5Hz, which is caused by the lightly damped zeros and
WP (12)
3.3s 1.2
0
poles in Pcom s as discussed in Section II. Consequently, the
According to the small gain theorem, the minimized H tracking bandwidth is limited to be 1.5Hz.
norm of Eq. (10) needs to be smaller than 1 to satisfy the three
design goals; otherwise it implies that one or more of the three 20
terms are violated. Since the trivial poles cause numerical
problem during the optimization process, the integrator in 0
*
Pcom s is replaced by a pole equal to 0.1. A solution Gc (Eq.
Magnitude(dB)
-20
(13)) is then obtained by standard robust synthesis which seeks
to find the smallest norm of Eq. (10). The resulting S , T , -40
T0*
GcS and their relationships with the design constraints are -60 1 / WI
shown in Figure 11. It can be seen that they all satisfy the Gc0*S*
design requirement and the tracking bandwidth is 2.18Hz, -80
which is higher than the frequency limitation of the original S0*
-100
system. 1 / WP
s 7 a6 s 6 a1s a0 -120
Gc (13) 10
-2
10
0
10
2
10
4
s 7 b6 s 6 b1s b0
Frequency(rad/sec)
where
[ a6 a5 a4 a3 a2 a1 a0 ]
[4.8e3 6.8e4 1.4e6 1.4e7 8.6e7 6.5e8 3.2e7]
FIGURE 12 ROBUST CONTROL DESIGN FOR Pcom s
0
[b6 b5 b4 b3 b2 b1 b0 ]
[1.6e3 3e4 6.3e5 8.2e6 6.2e7 5.3e8 4.7e8]
1.4
20
Step response of T*
0 1.2 Step response of T* 0
-20
Magnitude(dB)
1
Header height (m)
1.4
-40 T1_real
T2_real
0.8 1.2
-60 T*
1 / WI 1
Header height (m)
S* 0.6
-100 1 / WP 0.4
0.4
-120
-2 0 2 4 0.2
10 10 10 10 0.2
Frequency(rad/sec) 0
0 1 2 3
Time (sec)
4 5
0
0 10 20 30 40 50 60
FIGURE 11 ROBUST CONTROL DESIGN FOR P
*
com s Time (sec)
FIGURE 13 STEP RESPONSES COMPARISION