Untitled

Dead Time Compensation
Loops involving appreciable distance/velocity lags or dead time

arise from delays in signal transmission or from the distributed
parameter character of process elements.
Conventional 3-mode controllers, reviewed above, do not deal

satisfactorily with loops characterized by large dead times,
though the presence of derivative action provides for partial
stabilization of such loops.
The de-stabilizing effect of the presence of large dead times is

due to the contribution of dead time in increasing the phase lag
of the open loop transfer function, which means a reduction in
the cross over frequency.
The figure depicts qualitatively A.R.
the effect of the dead time (d) 1
on the Bode plot of a typical Gain Margin
Margin
Gain
open loop transfer function.
 d s
Ke
GH 
(1   1s)(1   2 s)

1 0
𝐺𝑎𝑖𝑛 𝑀𝑎𝑟𝑔𝑖𝑛 =
𝐴𝑅 𝜔𝑐𝑜
Margin
-90
Phase
𝑃ℎ𝑎𝑠𝑒 𝑀𝑎𝑟𝑔𝑖𝑛 = − 180 − 𝜑
𝐴𝑅=1 Phase Margin
-180
In Radians
Large d Small d
𝜋
𝑃ℎ𝑎𝑠𝑒 𝑀𝑎𝑟𝑔𝑖𝑛 = − 180 − 𝜑
𝐴𝑅=1 180
𝜑 = −𝑡𝑎𝑛−1 𝜏1 𝜔 − 𝑡𝑎𝑛−1 𝜏2 𝜔 −𝜔 𝜏𝑑
𝐴𝑅=1 𝐴𝑅=1 𝐴𝑅=1 𝐴𝑅=1
Sheet 1, Problem 3
10e  ds
3. Consider a first order process with dead time G p (s ) 
s1
which is controlled by a proportional controller with gain Kp.
Assume that the transfer functions of the measuring device and
control valve are respectively GH = GV = 1.
1. Show that as d increases, the ultimate gain of the
proportional controller decreases.
2. Assuming that d = 0.5 min, compute the gain of the
proportional controller so that the gain margin is 1.7 or the
phase margin is 30 which ever is safer.
3. Compute the steady state offset when the system is
subjected to a unit step change in the set point.
4. Design a dead time compensator for the system.
1. Show that as d increases, the ultimate gain of the proportional
controller decreases.
+
R C
-
The ultimate gain, Ku corresponds to co and at an AR = 1, thus
thus Eq. (1)
thus Eq. (2)

Eq. (1)
Eq. (2)
Differentiating Eq. (1): Which is always + ve
Differentiating Eq. (2):
Which is
always -ve,
thus
2. Assuming that d = 0.5 min, compute the gain of the proportional controller so that
the gain margin is 1.7 or the phase margin is 30 which ever is safer.
Eq. (1)
thus
Therefore For a gain margin of 1.7.
For a phase margin of 30:
At an AR = 1
Therefore the value of Kp= 0.22 is safer for the system to be stable
3. Compute the steady state offset when the system is subjected to a unit step change in
the set point. 10 𝑒 −0.5𝑠
+
R 𝐾𝑝 C
The closed loop transfer function: - 1+𝑠
10 𝑒 −0.5𝑠
𝐶 0.22 2.2 𝑒 −0.5𝑠
= 1+𝑠 =
𝑅 1 + 𝑠 + 2.2 𝑒 −0.5𝑠 1 + 𝑠 + 2.2𝑒 −0.5𝑠
1+𝑠
lim 𝑐(𝑡)
𝑡→∞ The offset is 1-0.6875=0.3125
1 2.2 𝑒 −0.5𝑠
= lim 𝑠𝐶 𝑠 = 𝑠 = 0.6875
𝑠→0 𝑠 1 + 𝑠 + 2.2𝑒 −0.5𝑠
Smith Dead Time Compensator
The shown loop is characterized by dead times in both the process
and measuring element transfer functions.
E P X
R + Gc Gv Gp e-1s C
-
B
GH e-2s
It may be simplified to the unity feedback configuration:

E’ C’ 1
R + Gc Gs e-ds C
- G H e   2s
C’
Where Gs = Gv Gp GH and d = 1 + 2
E’ C’ 1
R + Gc Gs e-ds C
- G H e   2s
C’
Where Gs = Gv Gp GH and d = 1 + 2
The open loop transfer function of the feed back loop is Gc Gs e-ds.
If (d) is large, the control loop is prone to instability (low cross
over frequency, co).
The addition of a feedback loop around the controller with

feedback elements having a special transfer function, the
parameters of which depend on the process transfer function
parameters, enables dead time elimination from the open loop
transfer function. This system is known as the “Smith Predictor”
or “Smith Dead Time Compensator”.
+ +  ds 1
R Gc G se Z
- - G H e   2s
(1  e  ds )G s
Which may be rearranged as
+ Gc Y e  1s
R Gs Z
-
1  G c G s (1  e ds ) GH
e  d s
The feedback loop control
ratio is given by: G cG s
 d s
Y 1  G c G s (1  e ) G cG s
  d s

R G cG se 1  G cG s
1  d s
1  G c G s (1  e )
Dead time is thus removed from the open loop transfer function
enabling easy tuning of the controller Gc. The system may be
represented by:
Y e-1s
R + Gc Gs Z
GH
It is recalled that loop outputs Y or C’ are not measurable, the

physically existing variables are C and Z.
It is noticed that the Smith predictor is based on a perfect

knowledge of the process and measuring element transfer
functions.
Errors in the parameters of the transfer function and/or variation
of the parameters during operation will result in incomplete dead
time compensation.
Sheet 1, Problem 3
4. Design a dead time compensator for the system.
The transfer function of the dead + + 10 𝑒 −0.5𝑠
R 𝐾𝑝 C
- - 1+𝑠
time compensator is
10 (1 − 𝑒 −0.5𝑠 ) 10 (1 − 𝑒 −0.5𝑠 )
1+𝑠 1+𝑠
𝐾𝑝 𝐾𝑝 (1 + 𝑠)
𝐺𝑐′ = =
10 𝐾𝑝 (1 − 𝑒 −0.5𝑠 ) 1 + 𝑠 + 10 𝐾𝑝 (1 − 𝑒 −0.5𝑠 )
1+ 1+𝑠
𝐾𝑝 (1 + 𝑠) 10 𝑒 −0.5𝑠 10𝐾𝑝 𝑒 −0.5𝑠

𝑂𝐿𝑇𝐹 = × =
1 + 𝑠 + 10 𝐾𝑝 (1 − 𝑒 −0.5𝑠 ) (1 + 𝑠) 1 + 𝑠 + 10 𝐾𝑝 (1 − 𝑒 −0.5𝑠 )
10𝐾𝑝 𝑒 −0.5𝑠
𝐶 1 + 𝑠 + 10 𝐾𝑝 (1 − 𝑒 −0.5𝑠 ) 10𝐾𝑝 𝑒 −0.5𝑠
= =
𝑅 1 + 𝑠 + 10 𝐾𝑝 (1 − 𝑒 −0.5𝑠 ) + 10𝐾𝑝 𝑒 −0.5𝑠 1 + 𝑠 + 10𝐾𝑝
1 + 𝑠 + 10 𝐾𝑝 (1 − 𝑒 −0.5𝑠 )
Sheet 3, Problem 1
+
X 𝐺1 𝑒 −𝜏1𝑠 + 𝐺2 𝑒 −𝜏2𝑠 C
It is required to control the process having the shown transfer function by a

measuring element, a control valve having the transfer functions (H) and (Gv)
respectively, and a controller having a transfer function (𝐺𝑐′ ) featuring dead time
compensation.
A. Sketch a block diagram of the conventional feedback loop.

B. Suggest the form of the transfer function of (𝐺𝑐′ )
C. Considering reference input changes only, show that the suggested transfer
function enables controller tuning on the basis of dead time elimination.
U
X +
+ +
R C
-
+
+ +
R C
-
+
+ + +
R C
- -
U
+ Y
+ + +
R C
- -
Considering U=0
Since the open loop transfer function does not contain dead
time, the controller can be tuned on the basis of dead time
elimination.

Untitled

Uploaded by

Copyright:

Available Formats

Untitled

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Untitled

Uploaded by

Copyright:

Available Formats

Dead Time Compensation

Loops involving appreciable distance/velocity lags or dead time

Conventional 3-mode controllers, reviewed above, do not deal

The de-stabilizing effect of the presence of large dead times is

The ultimate gain, Ku corresponds to co and at an AR = 1, thus

thus Eq. (1)

thus Eq. (2)

Differentiating Eq. (1): Which is always + ve

Differentiating Eq. (2):

For a phase margin of 30:

It may be simplified to the unity feedback configuration:

The addition of a feedback loop around the controller with

Which may be rearranged as

It is recalled that loop outputs Y or C’ are not measurable, the

It is noticed that the Smith predictor is based on a perfect

𝐾𝑝 (1 + 𝑠) 10 𝑒 −0.5𝑠 10𝐾𝑝 𝑒 −0.5𝑠

It is required to control the process having the shown transfer function by a

A. Sketch a block diagram of the conventional feedback loop.

You might also like