algorithm allows on/off solenoid valves to be used in place of costly servo ... Key Words : Pneumatic, Switching Control, On/off Solenoid Valve, Pulse Width ...
Journal of Mechanical Science and Technology, Vol t9, No. 2, pp 529 ~ 539, 2005
529
Intelligent Switching Control of Pneumatic Cylinders by Learning Vector Quantization Neural Network KyoungKwan Ahn*, ByungRyong Lee School o f Mechanical and Automotive Engineering, Umversay o f Ulsan, San 29, Muger2-dong, Nam-gu, Ulsan 680 749, Korea
The development of a fast, accm ate, and inexpensive position-controlled pneumatic actuator that may be applied to various practical positioning applications with various external loads ~s described in this paper A novel modified pulse-width modulatmn (MPWM) valve pulsing algorithm allows on/off solenoid valves to be used in place of costly servo valves A comparison between the system response of the standard PWM techmque and that of the modified PWM techmque shows that the performance of the proposed techmque was sigmficantly increased A state-feedback controller with positron, velomty and acceleration feedback was successfully ~mplemented as a contmuous controllel A switching algorithm fm control parameters using a learmng vector quantization neural network (LVQNN) has newly proposed, which classKies the external load of the pnemnatic actuator The effecUveness of this proposed control algorithm with smooth switching control has been demonstrated through experiments with various external loads
Key Words : Pneumatic, Switching Control, On/off Solenoid Valve, Pulse Width Modulatmn, Neural Network, Intelligent Control
Nomenclature U~mM (t) MPWM output Valve opening signal U~ ON duty ratio of valve tbr one tp
Um~x
Satm ated control input for MPWM
k
modulator Discrete sequence
1. Introduction
MPWM cycle Modified duty ratio with dead band region Modified duty ratio of on/off valve Y Yre: E
t~z T t
u(k)
Position of rodless cylinder Reference Position Maximum errol hmlt Dead time of valve MPWM cycle time Continuous time Sampled control Input of u ( t )
* Corresponding Author, E-mail kkahn@ulsan ackr IEL q-g2-52-259-2282,FAX ~82 5Z 259-2282 School of Mechamcal and AutomoUve Engineering, Umverstty of Ulsan, San 29, Muger2-dong, Nam-gu, Ulsan 680 749, Korea (Manuscript Received June 3, 2004, Revised November 17, 2004)
Pneumatm control systems play very important ioles m m d u s m a l automation systems owing to the advantages of low cost, easy maintenance, cleanliness, ready availability and cheap power source, etc (Anderson, 1967) A pamcularly welt-suited apphcatmn for pneumatic actuators ts the position control of 1obotic mampulators, loading/unloading systems, air balance systems and grippers, where stiff and lightweight StlUCtures are crltmal Unfoitunately, pneumatic actuators are subject to h,gh fiiCtlOn forces, deadband due to stlctlon, and dead-time due to the compresstblhty of air These nonhnearmes make accurate posmon control of pneumaUc actuators difficult to achieve As a result, a considerable amount of research
Copyright (C) 2005 NuriMedia Co., Ltd.
530
KyoungKwan Ahn and ByungRyong Lee
has been devoted to the development of various position control systems for pneumatic actuators (Noritsugu, 1995 ; Tang, 1995 ; Marchant, 1989 ; Kawamura, 1989). Many of these systems, though successful, use expensive proportional servo valves and pressure sensor feedback loops and the external loads are also assumed to be constant or slowly varying. The object of this paper is to implement inexpensive o n / o f f solenoid valves, rather than expensive servo valves, to develop a fast, accurate, inexpensive and intelligent pneumatic control system taking account of the changes of external loads. However, with solenoid valves, fine motion control is difficult to achieve because of the limitation of valve response time and its discrete o n / o f f nature. Previous studies (Noritsugu, 1986 ; Noritsugu, 1987 ; Muto, 1993 ; Mishra, 1994) have tried to implement o n / o f f solenoid valves for the position control of pneumatic actuators. Noritsugu (1995, 1996) has used the P W M method and solenoid valve to control the velocity and the position of pneumatic cylinders. Muto (1993) has used differential P W M (Pulse Width Modulation) method to control hydraulic actuators and Ahn and Tu (2004) has used intelligent switching control method in the position control of artificial muscle manipulator. These systems were successful in addressing smooth actuator motion in response to step inputs. However, some [imitations still exist, such as deterioration of the performance of transient responses due to cases of abrupt changes of external loads. To overcome this problem, the M P W M (modified pulse width modulation) o n / o f f valve control scheme and a switching control algorithm by LVQNN (Learning Vector Quantization Neural Network), have been newly proposed and control performance has been experimentally verified.
(Pentium i GHz) was applied to control the on./ off solenoid valve and to get experimental data. The stroke and the bore diameter of the rodless cylinder (SMC, MYIM32-1000L) were [000 mm and 32mm, respectively, and different masses could be attached to the table of the rodlcss cylinder. The displacement of the cylinder was measured by a linear scale (US Digital, resolution 0.05 mm) and the air pressure were also measured by air pressure sensor (SMC, ISE 40-01-22L) which was fed back to the computer through a 24 bit counter board (Advantech, P C L 833) and A / D board (Advantech, PCI 1731), respectively. The control signal was changed into pulse width modulated signal through our proposed MPWM algorithm. The control signal was sent to control the solid state relay and 8 pneumatic o n / o f f solenoid valves (MAC, 111B-872JD). The bandwidth of the valves used in the experiments was
,..........
1
i
i
i
i
Cl'mnt:,er I
h
]
Charnbeu ?
LJ
LL
/\
Fig, 1 Schematic diagram of pneumatic conlrol systern
2. Pneumatic Position Control System 2.1
Experimental apparatus
A schematic diagram of the control system is shown in Fig. 1. An IBM compatible computer
Copyright (C) 2005 NuriMedia Co., Ltd.
Fig. 2 Photograph of experimental apparatus
Intelligent Switching Conuol of Pneumatic Qvllnders by Learning Vector Quantization Neural Network 531 250 He. The experiments were conducted under
was longer than 4 ms.
a pressure of 0.5 MPa and all control software was coded in C. A photograph of the experi
2.3
mental apparatus was shown in Fig. 2.
When using on/off solenoid valves to control
Modified P W M
algorithm
the position of the cylinder, the control input u 2.2
Characteristics
of the o n / o f f solenoid
valve When using on/off solenoid valves to control
the position of a pneumatic cylinder, it is neces
must be converted into the pulse width modulation on/off signal in each solenoid valve. Previous studies have used 1) a conventional PWM scheme (Noritsugu, 1986, 1987; Muto, 1993),
sary to understand the characteristics of the o n / off solenoid valve9 This includes the dead time and the rise time of valve, the maximum current, etc. Figure 3 shows the current of the solenoid
where the valve opening time was proportional to
valve when the ON signal was applied to the o n / off solenoid valve for 100 ms. The current was measured from the voltage drop across the resist-
were considered. In modified PWM scheme, the dead-time of on/off solenoid valves is important in the design of our PWM algorithm because
ance, which is serially connected to the solenoid
the valve failed to move if the duration time of
valve. in the zoom graph of Fig, 3, the valve started to move alter t, and the valve was fully opened
ON signal to the solenoid valve was shorter than
after re. This means that the valve started to move when the duration of the ON pulse to the valve
the magnitude of the control input and 2) a modified PWM scheme (Shih, 1996, 1997), where the dead time of valve and other nonlinear terms
the dead-time of tile valve. To overcome this problem, we applied a modified PWM algorithm, where the minimum pulse width time was added to the output of the conventional PWM even if the control input was very small. However, the control input became zero within any allowable
Clllir
r
0.6
c"~'~l ~
"
6f-"-a~,
position error in order to prevent the oscillation of an actuator or a valve near the reference posit
J"P~'
tion. The limit of position error is expressed by r 0.4
3~
in equation 1.
>
3
O
0.2
r,
'i
i
N
b~"M(t'=l;
(k )T+tp(kl
