Professional Documents
Culture Documents
Weiyang Lin, Renhe Guan, Letian Yuan, Zhan Li & Mingsi Tong
To cite this article: Weiyang Lin, Renhe Guan, Letian Yuan, Zhan Li & Mingsi Tong (2018)
Position feedback dynamic surface control for pneumatic actuator position servo system,
Systems Science & Control Engineering, 6:1, 388-397, DOI: 10.1080/21642583.2018.1509400
Position feedback dynamic surface control for pneumatic actuator position servo
system
Weiyang Lina,b , Renhe Guanb , Letian Yuanb , Zhan Lib and Mingsi Tonga,b
a Key Laboratory of Micro-systems and Micro-structures Manufacturing of Ministry of Education, Harbin Institute of Technology, Harbin, China;
b Research Institute of Intelligent Systems and Control, Harbin Institute of Technology, Harbin, China
1. Introduction
adverse effects on the control performance of the system
With high dynamic performance, low pollution, safety (Schindele & Aschemann, 2009).
and strong environmental adaptability, the servo pneu- To improve accuracy, smoothness and rapidity of
matic systems are widely used in the industrial automa- the pneumatic system, many engineers and schol-
tion field. For instance, a pneumatic servo system ars have proposed considerable researches on pneu-
was designed for packing food products (Wang, Pu, matic position servo system to find the appropriate
& Moore, 1999). Servo pneumatic actuators were once and high-performance control method. For example,
applied in a climbing robot called Sky Cleaner IV (Zhang, a Proportion-Integration-Differentiation (PID) controller
Wang, & Zhang, 2009). Raparelli provided a pneumatic based on damping optimum was extended by a fric-
servo mechanism for the rehabilitation of the elbow joint tion compensation and the stabilization algorithm was
(Raparelli, Ivanov, & Palladino, 2017). Soft pneumatic actu- designed to achieve precise pneumatic position control
ators (SPAs) were utilized as fingers in robotic hands for in Situm, Pavkovic, and Novakovic (2004). A fuzzy logic
natural motion and light weight by Tao and Gu (2017). controller was implemented for controlling pneumatic
Toyama developed a flat-type pneumatic servo device for servo systems actuated by pneumatic muscle (Sándor,
mobile robots working in narrow spaces (Toyama & Waki- Szabolcs, & József, 2017). A Proportion-Differentiation
moto, 2016). An extensor pneumatic muscle actuator for (PD)-structured fuzzy-pulse width modulation controller
industries was implemented by Al-Ibadi, Nefti-Meziani, was developed using the feedback of the
and Davis (2017). piston position (Fathi & Najafi, 2017). Chiang and Chen
However, it is challenging to design a high- accomplished an intelligent adaptive control algorithm
performance tracking controller for the pneumatic posi- by using a neural network to minimize the tracking
tion servo system, because the precise parameters of error (Chiang & Chen, 2017). The paper (Dehghan,
the pneumatic system are difficult to be obtained (Ren, Taghizadeh, Surgenor, & Abu-Mallouh, 2011) explored
Gong, & Li, 2017). Besides, stick-slip of low-speed cylinder, the potential of a novel adaptive on-line neural net-
external force disturbance, noise interference and uncer- work compensator for the pneumatic position
tain nonlinear properties of pneumatic system also have control.
In general, the control of the pneumatic actuator posi- more, by constructing Lyapunov functions, stability con-
tion servo system has always been a complicated issue. ditions of these two methods are obtained, respectively.
There have been a variety of control strategies applied to In Section 4, the feasibility of our control method is
the pneumatic control including PID control, fuzzy con- verified by the experiment. From the results of the exper-
trol, neural network control,image-feature-based visual iment, it is obvious that compared to the traditional PID
servo control (Tong, Pan, Li, & Lin, 2018) and many other algorithm, the DSC method has better control perfor-
intelligent control methods. However, PID controller lacks mance and stronger robustness.
rigourous theoretical verification and its parameters can-
not be adjusted online. The control accuracy of fuzzy
2. The mathematical model of pneumatic
control is low and it requires specialized knowledge and
actuator
expert experience related to the system. The training
speed of neural network control is slow and it is also com- In order to control the pneumatic actuator, it is essential
plex to prove the stability and analyse convergence of to establish its mathematical model. The model can be
such control methods. Therefore, it is quite necessary for described as the following equations: force equilibrium
us to try new control methods to inspect whether these equation of the pneumatic actuator, mass flow equation
new methods can solve problems stated above. of cylinder chamber and pressure equation of cylinder
In this paper, the dynamic surface control (DSC) chamber. According to these equations, the open loop
of pneumatic system with position feedback is pre- transfer function of the pneumatic actuator is built. By
sented. DSC was proposed by Swaroop, Hedrick, Yip, controlling the input pressure of the cylinder and elimi-
and Gerdes (2000). By using the first-order filter after nating bad effects of contact force interference, the cylin-
each step of the backstepping control method, it over- der can output expected displacement.
comes the ‘explosion of terms’ problem caused by the
backstepping control method and reduces the amount
2.1. Force equilibrium equation of pneumatic
of calculation remarkably. DSC is effective in controlling
actuator
uncertain, mismatched nonlinear systems in the strict
feedback form. The pneumatic position servo system The schematic diagram of the pneumatic system is shown
exactly has many uncertain properties and external dis- in Figure 1. Consider the mass of moving platform as Mp ,
turbance, which can be possibly improved by the DSC the total mass of the end pneumatic actuator as m, the
method. output force of the pneumatic actuator as F1 , the force
The main contribution of this paper is stated as fol- between M and m as F2 , the contact force from workpiece
lows. Based on the theories about the pneumatic system, as FL and the angle between the actuator and vertical
a proper mathematic model on pneumatic position servo direction as α. The photograph of the actuator is shown
system is put forward in this paper. The model can be in Figure 2. As the photograph illustrates, the actuator is
applied in actual pneumatic control. Then, according to driven by the cylinder and it realizes reciprocating motion
the DSC theory, a dynamic surface controller and a mod- by changing the pressure of two chambers.
ified dynamic surface controller (MDSC) are designed, On the basis of Newton’s second law, force equilibrium
which can overcome the negative effects caused by the equation of pneumatic actuator can be obtained:
undesirable characteristics of pneumatic systems. The
F1 + F2 − FL + Gα − f − Kv ẋ = mẍ (1)
feasibility of the mathematical model and these designed
controllers can be proved by the following actual pneu- where Gα is the component of the gravity of m in the α
matic experiments. There is no doubt that the model and direction, which can be calculated by Gα = mg sin α, f is
controllers will give new reasonable solutions to many the Coulomb friction, x is the displacement of the cylinder
pneumatic control problems. rod and Kv is the viscous friction coefficient.
Specifically, this paper is organized as follows. In
Section 2, the pneumatic actuator designed by ourselves
and the gas flow in the cylinder are modelled. Then,
the open loop transfer function between input air pres-
sure and output displacement of cylinder rod is set up.
In Section 3, a dynamic surface controller is designed
for the pneumatic actuator based on the models estab-
lished in the previous section. Next, to eliminate mod-
elling uncertainties and resist force and noise interfer-
ence, a robust and adaptive DSC is proposed. What is Figure 1. Force analysis of the pneumatic system.
390 W. LIN ET AL.
Assume that the gas flow is continuous, it can be deduced Suppose that L is the length of stroke. The pressure
that the changing rate of gas mass flow is equal to the dif- differential equation of the cylinder can be obtained.
ference between the mass flow rate flowing in and out. According to the first law of thermodynamics, the energy
Let M = ρ · V denote the mass flow and the following equation of two cylinder chambers can be simplified as
equation is obtained: follows:
Assume that S1 P10 = S2 P20 , T10 = T20 = T0 and the pis- Ṁ1 = Kpa pc − KCa p1
(10)
ton only moves around the middle. Set the effective area Ṁ2 = Kpb pc − KCb p2
SYSTEMS SCIENCE & CONTROL ENGINEERING: AN OPEN ACCESS JOURNAL 391
F1 (s) = (ms2 + Kv s)x(s) + f (s) + FL (s) − Gα (s) − F2 (s) The main purpose of the DSC is to design a state feed-
(13) back control law u in order that output x tracks the desired
Since when the valve is open, pb is nearly constant trajectoryx1d .
about one standard atmospheric pressure. Therefore, To avoid ‘explosion of terms’ problem caused by back-
p2 (s) can be regarded as zero. Let stepping control, the DSC algorithm allows one virtual
control variable in backstepping control to pass through
βn = S21 P10 kr + S21 P10 kn + Kca Kv RT0 kr (14) a first-order filter and constructs another control variable.
Based on this method, the DSC controller is designed in
then, the transfer function A(s) between x(s) and pc (s)
the following.
can be obtained:
First step: design the surface function S1 = x1 − x1d ,
x(s) S1 RT0 k(Kcb + Kca r) then its dynamic function can be written as:
= (15)
pc (s) mV10 rs + (Kca RT0 kmr + Kv V10 r)s2 + βn s
3
pass through a first-order filter and then a new param- convenience, write ẋi as ẋi = xi+1 + fi (x1 , x2 , x3 ), where
eter variable x2d is produced, where τ2 is the filter time i = 1,2. When i = 3, ẋ3 = λu + f3 (x1 , x2 , x3 ), where λ =
constant. (K1 /A). Then simplify zi żi . According to the property of the
τ2 ẋ2d + x2d = α2 (23) filter, the following equation can be attained:
Third step: the last surface function is S3 = x3 − x3d , and where φi+1 (Si , xi ) = ki Ṡi + ((∂fi )/(∂xk ))ẋk − ẍid . Besides,
the control law is zi żi can be represented as
1 C B 2
u= −k3 S3 + x2 + x3 + ẋ3d (27) zi+1
λ A A zi+1 · żi+1 = − + zi+1 · φi+1 (Si , xi ) (34)
τi+1
so far, the dynamic surface controller has already been
designed. then Equation (29) can be transformed as
z22 z2
V̇(Si , zi ) = − k1 S21 − k2 S22 − k3 S23 − − 3
3.2. Lyapunov stability verification of DSC controller τ2 τ3
Define filtering variable z2 = x2d − α2 = −τ2 ẋ2d and sim- + S1 z2 + S1 S2 + S2 z3 + S2 S3 + z2 φ2 + z3 φ3
ilarly filtering variable z3 = x3d − α3 = −τ3 ẋ3d . Then, the (35)
Lyapunov function of the dynamic surface controller is Assume that V(0) is the initial state of the Lyapunov
constructed, function and p is an upper bound of it. It is obvious that
xid and its all-order derivative are bounded, because xid is
1 2 1
V(Si , zi ) = (S1 + S22 + S23 ) + (z22 + z32 ) (28) the desired trajectory set by ourselves. Then, consider the
2 2
following two compact sets 1 and2 :
When the function V is equal to zero, the actual and
expected motion state of the object is static. However, 1 = {(xid , ẋid , ẍid ) : (xid )2 + (ẋid )2 + (ẍid )2 < ζ }
the object cannot be static all the time in actual circum- (36)
2 = {(Si , zi ) : S21 + S22 + S23 + z22 + z32 < 2p}
stances. Therefore, the function V is larger than zero. Next,
take the derivation of Equation (28), If the function φi belongs to the compact set {1 ×
2 }, it has the upper bound which can be set as Mi .
V̇(Si , zi ) = S1 Ṡ1 + S2 Ṡ2 + S3 Ṡ3 + z2 ż2 + z3 ż3 (29)
According to Young’s inequality,
In order to obtain the negative derivative of the Lya-
punov function, it is necessary to simplify Si Ṡi and zi żi . S1 z2 + S1 S2 + S2 z3 + S2 S3 + z2 φ2 + z3 φ3 ≤ S21 +
When i = 1,2, Si Ṡi can be simplified as: 3 2 1 2 1 2 1 2 1 2 2 1 2 2 (37)
S2 + S3 + z2 + z3 + z2 φ2 + z3 φ3 + 1
2 2 2 2 2 2
Si Ṡi = Si (ẋi − ẋid )
therefore,
= Si (xi+1 − αi+1 − ki Si )
d 3
= Si (zi+1 − xi+1 + xi+1 − ki Si ) (30) V̇(Si , zi ) ≤S21 (−k1 + 1) + S22
−k2 +
2
= Si (zi+1 + Si+1 − ki Si )
2 1 2 1 1 1 2
+ S3 −k3 + + z2 − + + φ2
= Si zi+1 + Si Si+1 − ki S2i 2 τ2 2 2
particularly, when i = 3, S3 S˙3 = −k3 S23 , z4 = 0, α4 = 0, 1 1 1
+ z32 − + + φ32 (38)
x4d = 0 and x4 = 0, since the system is the third order. For τ3 2 2
SYSTEMS SCIENCE & CONTROL ENGINEERING: AN OPEN ACCESS JOURNAL 393
20 20
PID PID
DSC DSC
10 10
Sine Sine
y/cm
y/cm
0 0
-10 -10
-20 -20
0 0.5 1 0 0.5 1
t/s t/s
select b > 0, k1 ≥ 1 + b, k2 ≥ 32 + b, k3 ≥ 12 + b, (1/τ2 ) ≥ 1.6 Hz sine signal and the right one is about 5 Hz sine sig-
(M22 /2) + 12 + b, (1/τ3 ) ≥ (M23 /2) + 12 + b, then nal. Besides, the results of step input and ramp input are
shown in Figure 4.
M22 φ22 M2 M23 φ32 M2 From the results, the traditional PID controller can
V̇ ≤ −2bV + 1 + − 2 z22 + − 3 z32 track the sine signal approximately when the frequency
2M22 2 2M23 2
of the input sine is 1.6 Hz while it cannot track the sine sig-
(39)
nal with 5 Hz frequency. However, the DSC controller can
Since V(0) ≤ p and Mi is the upper bound of φi , the lat- accomplish the task of tracking the input sine signals with
ter two items are not greater than zero. Set b > (1/2p), 1.6 Hz and 5 Hz, which proves that it has a greater band-
then, V̇ < 0 can be easily verified. Therefore, the function width than the traditional PID controller. When it comes
V is smaller than p and it also belongs to the compact sets. to step signal and ramp signal, the DSC controller has a
Thus, if the related parameters in the former equations faster response speed, smaller steady-state error and less
satisfy the certain conditions, the control method can be overshoot.
stable.
1.5 3
PID PID
DSC Ramp
2
1 Step DSC
y/cm
y/cm
1
0.5
0
0 -1
0 0.5 1 1.5 2 0 0.5 1 1.5 2
t/s t/s
According to the control law and the definition of choose c > 0, k1 ≥ 2 + c, k2 ≥ 94 + c, k3 ≥ 14 + c, (1/τi ) ≥
estimation error, (M2i /2) + 14 + c, ζi = (2c/ξi ), ηβ = (2c/γβ ) and ηfw =
u
λu = u + λβ (41) (2c/γfw ). Thus,
20
Sine 1 Step
MDSC MDSC
10 0.8
DSC DSC
0.6
y/cm
y/cm
0
0.4
-10
0.2
-20 0
0 0.5 1 0 0.5 1 1.5 2
t/s t/s
30 1.5
Sine Step
20 MDSC MDSC
DSC 1 DSC
10
y/cm
y/cm
0
0.5
-10
-20 0
0 0.5 1 0 0.5 1 1.5 2
t/s t/s
y/cm
0
0.6
-5
0.4
0.2 -10
0 -15
0 0.2 0.4 0.6 0.8 1 0 0.5 1 1.5 2
t/s t/s
From Figure 7, it can be deduced that both two con- 4.1. Brief introduction about the experimental
trollers can make the pneumatic system track input sig- platform
nals and have not bad control performance. Besides, the
The whole pneumatic system is controlled by DSP and
MDSC controller designed by us can accomplish the con-
driven by cylinders. The pressure of intake chamber in
trol tasks of pneumatic systems like other published ones.
the cylinder determines the output position of the sys-
tem. By controlling the voltage of proportional pressure
4. Experiment on pneumatic compliant system regulator, the intake chamber pressure can be adjusted
In this section, two pneumatic experiments are designed to regulate the system output. In addition, the pressure
to verify the accuracy of previous theories. One is the posi- of two chambers in the cylinder can be collected by the
tion step response experiment for comparing the control proportional pressure regulator. In order to obtain posi-
performance among the traditional PID controller, nor- tional information, a position sensor is used. What is more,
mal DSC controller and the MDSC controller. The other a two-position and five-port electronic valve is applied
is the force impulse experiment to test whether MDSC to ensure that the cylinder piston can reciprocate. The
controller has stronger robustness than normal DSC con- experimental platform is shown in Figure 8 and the struc-
troller as simulation results show. ture of experimental platform is shown in Figure 9.
396 W. LIN ET AL.
1
DSC
0.8 MDSC
0.6
y/cm
0.4
0.2
0 0.5 1 1.5
t/s
Figure 8. Experimental platform with industrial robot. Figure 11. Result of the force impulse experiment.