che302 lecture ix frequency responses · 2001-11-27 · che302 process dynamics and control korea...
TRANSCRIPT
CHE302 Process Dynamics and Control Korea University 9-1
CHE302 LECTURE IXFREQUENCY RESPONSES
Professor Dae Ryook Yang
Fall 2001Dept. of Chemical and Biological Engineering
Korea University
CHE302 Process Dynamics and Control Korea University 9-2
DEFINITION OF FREQUENCY RESPONSE
• For linear system– “The ultimate output response of a process for a sinusoidal
input at a frequency will show amplitude change and phaseshift at the same frequency depending on the processcharacteristics.”
– Amplitude ratio (AR): attenuation of amplitude,– Phase angle ( ): phase shift compared to input– These two quantities are the function of frequency.
ProcessInput Output
sinA tω ˆsin( )A tω φ+
After all transienteffects are decayed out.
t
φ
( )u t
( )y t
-A
AA
A−
φ
ˆ/A A
CHE302 Process Dynamics and Control Korea University 9-3
BENEFITS OF FREQUENCY RESPONSE
• Frequency responses are the informativerepresentations of dynamic systems– Audio Speaker
– Equalizer
– Structure
Elec. Signal Sound wave
Rawsignal
Processedsignal
Expensivespeaker
Cheapspeaker
Old Tacoma bridge New Tacoma bridge
Old Tacomabridge
New Tacomabridge
Adjustablefor each
frequencyband
CHE302 Process Dynamics and Control Korea University 9-4
– Low-pass filter
– High-pass filter
– In signal processing field, transfer functions are called “filters”.
CHE302 Process Dynamics and Control Korea University 9-5
• Any linear dynamical system is completelydefined by its frequency response.– The AR and phase angle define the system completely.– Bode diagram
• AR in log-log plot• Phase angle in log-linear plot
– Via efficient numerical technique (fast Fourier transform,FFT), the output can be calculated for any type of input.
• Frequency response representation of a systemdynamics is very convenient for designing afeedback controller and analyzing a closed-loopsystem.– Bode stability– Gain margin (GM) and phase margin (PM)
CHE302 Process Dynamics and Control Korea University 9-6
• Critical frequency– As controller gain changes, the amplitude ratio (AR) and the
phase angle (PA) change.– The frequency where the PA reaches –180° is called critical
frequency ( ).– The component of output at the critical frequency will have the
exactly same phase as the signal goes through the loop due tocomparator (-180 °) and phase shift of the process (-180 °).
– For the open-loop gain at the critical frequency,• No change in magnitude• Continuous cycling
– For• Getting bigger in magnitude• Unstable
– For• Getting smaller in magnitude• Stable
cω
( ) 1OL cK ω =
RGc … GP
+
-
YE
Signchange
Signchange
( ) 1OL cK ω >
( ) 1OL cK ω <
CHE302 Process Dynamics and Control Korea University 9-7
• Example– If a feed is pumped by a peristaltic pump to a CSTR, will the
fluctuation of the feed flow appear in the output?
– V=50cm3, q=90cm3/min (so is the average of qi)• Process time constant=0.555min.
– The rpm of the peristaltic pump is 60rpm.• Input frequency=180rad/min (3blades)
– The AR=0.01
t
qi
V
cAi, qi
Peristaltic pump cA, q
If the magnitude of fluctuation of qi is 5% of nominalflow rate, the fluctuation in the output concentrationwill be about 0.05% which is almost unnoticeable.
( constant)Ai Ai A
dcV q c qc q
dt= − ≈
/( )( ) ( / ) 1
Ai AiA
i
C C qC sq s Vs q V q s
= =+ +
( 100)ωτ =
CHE302 Process Dynamics and Control Korea University 9-8
OBTAINING FREQUENCY RESPONSE
• From the transfer function, replace s with
– For a pole, , the response mode is .– If the modes are not unstable ( ) and enough time elapses,
the survived modes becomes . (ultimate response)
• The frequency response, is complex as afunction of frequency.
jω
( ) ( )s jG s G jω ω=→Transfer function Frequency response
( )G jω
( ) Re[ ( )] Im[ ( )]G j G j j G jω ω ω= +
2 2( ) Re[ ( )] Im[ ( )]AR G j G j G jω ω ω= = +
( )1( ) tan Im[ ( )]/ Re[ ( )]G j G j G jφ ω ω ω−= =R
s jα ω= + ( )j te α ω+
0α ≤j te ω
Re
Im
φ
( )G jω
AR
ω ↑
Nyquistdiagram
Bode plot
CHE302 Process Dynamics and Control Korea University 9-9
• Getting ultimate response– For a sinusoidal forcing function– Assume G(s) has stable poles bi.
– Without calculating transient response, the frequency responsecan be obtained directly from .
– Unstable transfer function does not have a frequency responsebecause a sinusoidal input produces an unstable outputresponse.
2 2( ) ( ) AY s G ss
ωω
=+
12 2 2 2
1
( ) ( ) n
n
A Cs DY s G ss s b s b s
αω α ωω ω
+= = + + ++ + + +
L
( ) ( ) D CG j A Cj D G j j R jIA A
ω ω ω ω ω= + ⇒ = + = +
Decayed out at large t
ˆ, ( cos sin ) sin( )ulC IA D RA y A I t R t A tω ω ω φ= = ⇒ = + = +
2 2 1ˆ/ = ( ) and tan ( / ) ( )AR A A R I G j I R G jω φ ω−∴ = = + = =R
( )G jω
CHE302 Process Dynamics and Control Korea University 9-10
• First-order process
• Second-order process
( )( 1)
KG s
sτ=
+
2 2( ) (1 )(1 ) (1 )
K KG j j
jω ωτ
ωτ ω τ= = −
+ +
2 2
1( )1
NAR G jωω τ
= =+
1( ) tan ( )G jφ ω ωτ−= = −R
2 2( )
( 2 1)K
G ss sτ ζτ
=+ +
2 2( )
(1 ) 2K
G jj
ωτ ω ζτω
=− +
2 2 2 2( )
(1 ) (2 )
KA R G jωω τ ζ ω τ
= =− +
1 12 2
Im( ( )) 2( ) tan tanRe( ( )) 1
G jG jG j
ω ζωτφ ωω ω τ
− −= = = −−
R
1ζ >
Corner freq.
CHE302 Process Dynamics and Control Korea University 9-11
• Process Zero (lead)
• Unstable pole
( ) 1aG s sτ= +
( ) 1 aG j jω ωτ= +
2 2( ) 1N aAR G jω ω τ= = +
1( ) tan ( )aG jφ ω ωτ−= =R
1( )
( 1)G s
sτ=
− +
2 2
1 1( ) (1 )
1 1G j j
jω τω
τω τ ω= = +
− +
2 2
1( )1
A R G jωω τ
= =+
1 1Im( ( ))( ) tan tanRe( ( ))
G jG jG j
ωφ ω ωτω
− −= = =R
CHE302 Process Dynamics and Control Korea University 9-12
• Integrating process
• Differentiator
• Pure delay process
1( )G s
As=
1 1( )G j j
jA Aω
ω ω= = −
1( )NAR G jA
ωω
= =
1 1( ) tan ( )0 2
G j πφ ωω
−= = − = −⋅
R
( ) sG s e θ−=
( ) cos sinjG j e jθ ωω θω θω−= = −
( ) 1AR G jω= =1( ) tan tanG jφ ω θω θω−= = − = −R
( )G s As= ( )G j jAω ω=
( )NAR G j Aω ω= =
1 1( ) tan ( )
0 2G j
πφ ω
ω−= = =
⋅R
CHE302 Process Dynamics and Control Korea University 9-13
SKETCHING BODE PLOT
• Bode diagram– AR vs. frequency in log-log plot– PA vs. frequency in semi-log plot– Useful for
• Analysis of the response characteristics• Stability of the closed-loop system only for open-loop stable
systems with phase angle curves exhibit a single critical frequency.
1 2 3
( ) ( ) ( )( )( ) ( ) ( )
a b cG s G s G sG sG s G s G s
= LL 1 2 3
( ) ( ) ( )( )( ) ( ) ( )
a b cG j G j G jG jG j G j G j
ω ω ωωω ω ω
= LL
1 2 3
( ) ( ) ( )( )
( ) ( ) ( )a b cG j G j G j
G jG j G j G j
ω ω ωω
ω ω ω=
LL
1 2 3
( ) ( ) ( ) ( )( ) ( ) ( )
a b cG j G j G j G jG j G j G j
ω ω ω ωω ω ω
= + + +− − − −
R R R R LR R R L
CHE302 Process Dynamics and Control Korea University 9-14
• Amplitude Ratio on log-log plot– Start from steady-state gain at =0. If GOL includes either
integrator or differentiator it starts at or 0.– Each first-order lag (lead) adds to the slope –1 (+1) starting at
the corner frequency.– Each integrator (differentiator) adds to the slope –1 (+1)
starting at zero frequency.– A delays does not contribute to the AR plot.
• Phase angle on semi-log plot– Start from 0° or -180° at =0 depending on the sign of steady-
state gain.– Each first-order lag (lead) adds 0° to phase angle at =0, adds
-90° (+90°) to phase angle at = , and adds -45° (+45°) tophase angle at corner frequency.
– Each integrator (differentiator) adds -90° (+90°) to the phaseangle for all frequency.
– A delay adds – to phase angle depending on the frequency.
∞
∞
ω
ω
ω
θω
ω
CHE302 Process Dynamics and Control Korea University 9-15
Examples
1. 2.( )(10 1)(5 1)( 1)
KG s
s s s=
+ + +
0.55(0.5 1)( )(20 1)(4 1)
ss eG ss s
−+=+ +
G1
G2 G3
G4G5
G1 G2 G3
CHE302 Process Dynamics and Control Korea University 9-16
3. PI: 5. PID:
4. PD:
1( ) 1CI
G s Ksτ
= +
1( ) 1C DI
G s K ss
ττ
= + +
Slope=-1 Slope=+1
1/ at 0Notch I Dω τ τ φ= = °1/ at 45b Iω τ φ= = − °
Slope=-1
( )( ) 1C DG s K sτ= +Slope=+1
1/ at 45b Dω τ φ= = − °
CHE302 Process Dynamics and Control Korea University 9-17
NYQUIST DIAGRAM
• Alternative representation of frequency response• Polar plot of ( is implicit)
– Compact (one plot)– Wider applicability of stability
analysis than Bode plot– High frequency characteristics will be
shrunk near the origin.• Inverse Nyquist diagram: polar plot of
– Combination of different transfer function components is noteasy as with Nyquist diagram as with Bode plot.
Re
Im
φ
( )G jω
AR
ω ↑
Nyquistdiagram( ) Re[ ( )] Im[ ( )]G j G j j G jω ω ω= +
( )G jω ω
1/ ( )G jω