Professional Documents
Culture Documents
KEYWORDS
Control systems, DC drive, fuzzy control, model,
MATLAB, NCD Blockset, Power System Blockset dI/dt
f(u) 1/s 1
Integrator1 current
ABSTRACT voltage Mux
1 dw/dt
An industrial DC drive (22kW) with fuzzy controller f(u) 1/s 2
is simulated. Two models (linear and nonlinear) and 2 velocity
Integrator2
two controllers (PID and fuzzy) are investigated. Using load
fuzzy controller for DC drive operation was successful. Fig. 1 Linear model of DC motor
(custom subsystem MotorDC)
INTRODUCTION
Converter/rectifier is described as first order inertia
Two mathematical models of a DC drive are used. The kp(s)
first model is build as linear transfer function of Gconv = --------------
converter and DC motor. The second model is build Tmip·s +1
using advanced blocks from Power System Blockset where:
(PSB) library. The library is an extension of kp -gain of converter/rectifier
MATLAB/Simulink environment from The Tmip -mean dead time of converter/rectifier
MathWorks Inc. Using fuzzy logic and PSB library
model seems to be new and promising approach to The dead time Tmip may vary from zero to one-half the
control of an electric drive. period of an AC source (0.01s for 50 Hz) [Devan,
1984]. It is assumed that six-phase thyristor bridge
LINEAR MODEL OF DC DRIVE with mean dead time Tmip=1.67ms is used in the
converter.
A linear and nonlinear model of DC drive will be used.
The linear model consists of two parts: A classic DC drive with two PID controllers is
converter/rectifier and DC motor. A linear model of presented on figure 8. It was assumed to neglect a
DC motor (figure 1) was build using Simulink blocks. derivative signal and to use PI operation of a current
There are two inputs (voltage and load) and two controller only. Parameters of the current controller
outputs (angular motor velocity and current). Its were derived from the model parameters using rules of
parameters are computed automatically from nominal module and symmetry. Then Nonlinear Control
catalogue data: motor power, voltage, current, speed, Design Blockset (NCD) was used for automatic tuning
etc). It is very convenient to use nominal motor data as of the controller parameters (fig.2), to minimise the
rotor inductance and resistance. DC motor constant transient overshot.
and other internal motor parameters are difficult to
find.
186
USING POWER SYSTEM BLOCKSET TO
MODEL THE DC DRIVE
60 20
10
40
0
20
-10
0 0.5 1 1.5
0 Time
0 0.5 1 1.5 2 2.5
-10 60
0 0.5 1 1.5 2 2.5
Time
40
Fig. 3 Current (upper red) and angular velocity
(lower blue) of DC motor. Simulink and linear model 20
were used for simulation.
0
Similar procedure was used to find parameters of 0 0.02 0.04 0.06 0.08 0.1
velocity controller. The simulation results (DC motor Fig. 5. Detail of simulated current signal using
current and angular velocity vs. time) are presented on Simulink and Power System Blockset
figure 3. This is raw simulation as linear model has
very low granularity: AC component of current and The three-phase bridge converter is the most frequently
switching of currents in thyristor bridge are neglected. used motor control system. Two of six thyristors
Only envelope of transients can be seen on simulation conduct at any time instant. Gating of each thyristor
output. initiates a pulse of load current; therefore this is a six-
187
pulse controlled rectifier. The three-phase six-pulse
rectifier is also capable of inverter operation in the
fourth quadrant.
Electrical phenomena of thyristor bridge and DC motor
are modelled very exactly. Simulation results (figure 4
and 5) are almost exact with real measurement data 5
on industrial object, but computation is slow
0
ust
comparing to linear model.
-5 10
FUZZY CONTROLLER OF DC DRIVE
0
The fuzzy controller is presented on figure 10. 1
0 -10 error
Advanced model using Power System Blockset is used, INTEGerror -1
but transfer function model can also be useful for Fig. 6. Fuzzy control surface
preliminary tuning of controller parameters.
<Ia>
Linguistic variables and rules 80
60
There are two fuzzy variables (error and INTEG error)
and seven linguistic variables (from big negative to 40
big positive). The fuzzy controller attributes are:
20
type: 'mamdani'
andMethod: 'prod' 0
orMethod: 'max' 0 0.02 0.04 0.06 0.08 0.1
time
defuzzMethod: 'centroid'
impMethod: 'prod' Fig. 7. Detail of simulation results (motor current).
aggMethod: 'max' Fuzzy controller (here) react faster than PI (see fig.
input: [1x2 struct] 5)
output: [1x1 struct]
rule: [1x25 struct] FINAL REMARKS
The membership functions (pimf and gausmf are Both simple transfer function model and the advanced
used)and rules are design tools that give opportunity set of linear and nonlinear blocks from Power System
to model a control surface and controller properties. It Blockset are useful for tuning the DC drive control
is obvious that using this attributes one can more system. Advanced models build with Power System
precisely fulfil a quality criterion in full operational Blockset blocks are suitable for preliminary
range. The control surface (figure 6) is defined with 25 verification of control system, as AC component of
rules. current and switching phenomena of thyristor bridge
are not neglected. The fuzzy controller is more difficult
to design comparing with PID controller but has more
RESULTS AND CONCLUSION
design parameters and is more suitable to fulfil
Simulation output for fuzzy controller is similar to PI nonlinear quality criterion in all operational range – as
controller output presented on figure 4 – unless one seen on figure 6. For real time operation a discrete
consider how controller react for external disturbation. fuzzy algorithm can be implemented on
The investigation showed that even simple fuzzy microcomputer, DSP or ASIC chip, which is more
controller used to control DC drive operation (fig. 10) suitable for industrial application.
is more precise and faster than of PI controller
(compare fig. 7 with fig. 5). ACKNOWLEDGEMENTS
This work was supported by Cracow University of
Technology, grant F-3/147/DS/2000.
188
Run Reg_N
inidcdr 22kW
Reg_I
Speed nz*
RI-RN
reference Ia* 1 + ia-ws1
regPI ust kp(s)
sp_ref
SP_ramp
tfi.s+1 - regPI
ust <Ia>
Tmip.s+1 MotorDC
Controller_n filtr
Controller_I converter <ws>
MotorDC
ki
kom
kw
A Jobl=2.7
i
K +
- A+
B
w
i1
C TL
A
pulses Load torque 0 w0
A-
RN RI
Speed nz*
reference ia* 1 + ust
regPI
sp_ref SP_ramp tfi.s+1 - regPI
Control ler_n fil tr
Controller_I
8.0
Run
dcdrpar 22kW kia
ki
kom
kw
189
A Jobl=2.7
i
K +
- A+
B
w
i1
C TL
A
pulses Load torque A-
0 w0
synchronization signals 6 - pulse
w0
Converter DC_Motor
+
v
-
AC
vac
+ Ia(A) & w
v BA
-
Pout
vba
CB
+
v
-
vcb CTL
Va1 Vb Vc Pulse_generator
kia
ki
RN
RI
Speed nz*
1 e
reference regPI ia*
ust
sp_ref tfi.s+1 Mux
SP_ramp e 1 INTEGerror
Controller_n filtr
Sum s
Integrator
Fuzzy
Controller_I
Run 8.0
inidcdr 22kW
kom
kw
190