Upload
atanu-maiti
View
214
Download
0
Embed Size (px)
Citation preview
8/16/2019 02 - PID Training - Process Dynamics
http://slidepdf.com/reader/full/02-pid-training-process-dynamics 1/27
PROCESS CONTROL
THEORYFUNDAMENTAL PRINCIPLES
8/16/2019 02 - PID Training - Process Dynamics
http://slidepdf.com/reader/full/02-pid-training-process-dynamics 2/27
• All Processes are dynamic
• i.e. they change with time.
• If a plant were totally static with respect to all it’s variables, then control
would be easy
• But in reality the dynamics are constantly changing
• We will define the basic terms and show how they relate to process response
• The change with respect to time of the process variables such astemperature, pressure, flow, composition etc, due to
• controlled changes e .g. feed rate, temperature, pressure etc
• uncontrolled changes e.g. ambient temperature, feed composition etc
• Understanding process dynamics is fundamental for achieving good control
PROCESS DYNAMICS
8/16/2019 02 - PID Training - Process Dynamics
http://slidepdf.com/reader/full/02-pid-training-process-dynamics 3/27
• Process Response
• Process Gain
• Process Deadtime
• Process Lag
• Order
• Linearity
• Non-self regulating systems
• Plant tests
PROCESS DYNAMICSTERMS
8/16/2019 02 - PID Training - Process Dynamics
http://slidepdf.com/reader/full/02-pid-training-process-dynamics 4/27
Feed flow rate
Fuel gas valve position
Time
Time
Coil Outlet temperature
Fuel gas
PID OPEN LOOP
8/16/2019 02 - PID Training - Process Dynamics
http://slidepdf.com/reader/full/02-pid-training-process-dynamics 5/27
• Dynamic response of a process can usually be characterized by 3parameters :
• Process Gain
• Kp or G and is the Change in the process variable divided by thechange in the manipulated variable. Expressed in Engineering units
• Deadtime
• DT or !, the time between MV changing and a noticeable changein PV. Expressed in minutes
• Lag
• T1 or " Effects the rate at which the PV responds to an MVchange. Expressed in minutes
PROCESS RESPONSECHARACTERIZATION
8/16/2019 02 - PID Training - Process Dynamics
http://slidepdf.com/reader/full/02-pid-training-process-dynamics 6/27
• Laplace Transforms are simply amathematical technique which can expressequations in the time domain
• This allows straight-forward calculations tobe carried out instead of solving complexdifferential equations
• Format is easily understood
• Let me illustrate by some examples
LAPLACETRANSFORM
Simple Laplace transform techniques shall be detailed here. L(f(t)) = !f(t)e-stdt ; explain simple di" erential equations.Provide introduction to state space representation. Also talk on block diagram and its reductionExample on few simple laplace: A, e-at, t, F’Example on block diagram reduction
8/16/2019 02 - PID Training - Process Dynamics
http://slidepdf.com/reader/full/02-pid-training-process-dynamics 7/27
LAPLACE FORMAT - 1ORDER
Process Gainof 3.5
Process lag is 1storder and 20 min
Process Deadtimeof 5 min
Laplace on control domain - I order di" erential equation.#dC/dt + Cin = CoutLaplace of this becomesCout = Cin/(1+#s)
8/16/2019 02 - PID Training - Process Dynamics
http://slidepdf.com/reader/full/02-pid-training-process-dynamics 8/27
Process Gainof 0.5 Process lag is 2nd
order and consists of2 lags each 10 min long
Process Deadtimeof 2 min
Process Leadof 5 min
( )( ) s
e
s s
s 2
110110
155.0
!
++
+
LAPLACE FORMAT - IIORDER
If the 2 time constants di" er much, then this process can be assumed as I order.
8/16/2019 02 - PID Training - Process Dynamics
http://slidepdf.com/reader/full/02-pid-training-process-dynamics 9/27
Process Gainof 20 Process lag is 2nd
order and consists of2 lags each 10 min long(alternative representation)
No Process Deadtime
Process Leadof 0.5 min
s
e
s s
s 0
2120100
15.00.20
!
++
+
LAPLACE FORMAT - IIORDER
Talk on laplace of P, I and D controllers.For P, G(s) = Kc = Ps/EsFor P&I, G(s) = Kc(1+1/"1s) = Ps/EsFor P&D, G(s) = Kc(1+ "Ds) = Ps/EsFor PID, G(s) = Kc(1+1/"1s+ "Ds) = Ps/EsExplain on closed loop block diagramC/R = GcGp/(1+GcGp) - basis for PID loop tuning
8/16/2019 02 - PID Training - Process Dynamics
http://slidepdf.com/reader/full/02-pid-training-process-dynamics 10/27
At time, t, h = height of liquid in vessel
h = H exp (-t/ !)
When t = !
h = H.exp(-1) = 0.368 H
i.e. level has fallen by (1-0.368) = 0.632 H
Time
H
H1
1 ORDER SYSTEM
8/16/2019 02 - PID Training - Process Dynamics
http://slidepdf.com/reader/full/02-pid-training-process-dynamics 11/27
Time
H
Time
H Notice Apparent deadtime
Top tank is 1st order as before
2
1
Critically damped system ("1 = "2)
II ORDER SYSTEMEQUAL TIME CONSTANTS
8/16/2019 02 - PID Training - Process Dynamics
http://slidepdf.com/reader/full/02-pid-training-process-dynamics 12/27
Over damped system ("1 > "2)
Time
H
Time
H Looks like 1st order
Top tank is much slower
t2 is the dominant time constant
SLOW
FAST2
1
II ORDER SYSTEM
#1 > #2
8/16/2019 02 - PID Training - Process Dynamics
http://slidepdf.com/reader/full/02-pid-training-process-dynamics 13/27
Under damped system ("1 < "2)
Time
H
Time
H Notice inverse response, level actually rises
initially. This confuses a simple feedback
controller system
Imagine Top tank now empties very fast
FAST
SLOW2
1
II ORDER SYSTEMT1 < T2
Explain typical underdamped process response and its characterOvershoot = exp (-$% / & 1- %2)Settling time = 4"/ % - 98%Decay ratio = OS2
8/16/2019 02 - PID Training - Process Dynamics
http://slidepdf.com/reader/full/02-pid-training-process-dynamics 14/27
Time
H
No. 6 Looks like
DT+1st order Lag!
due to series of lags
1
2
3
4
6
5
MANY TANKS INSERIES
8/16/2019 02 - PID Training - Process Dynamics
http://slidepdf.com/reader/full/02-pid-training-process-dynamics 15/27
First order with Deadtime can approximate most of the processFirst order model approximation of 20th order system is shown below
Time
20th order system
Deadtime + 1st order lag
model of system
I ORDER + DT
What does a real plant response look like?
Real plants exhibit very high order dynamics (i.e. many lags in series) We can approximate a high order system such as the 20 order one above as a 1st order system with
deadtime with a pretty good fit
Work out an example: Y/X = 3/(0.1s+1)(0.5s+1)(s+1)(3s+1)reduced form: Y/X = 3exp(-1.6s)/(3s+1)plot the response and try compare it.
Pade approximation for dead time shall also be explained here: 1 order: (1+tds/2)/(1-tds/2)Example:Y/X = exp(-3s)/(10s+1)
with first order pade, it is = (1.5s+1)/(10s+1)(1-1.5s), plot the response and see
8/16/2019 02 - PID Training - Process Dynamics
http://slidepdf.com/reader/full/02-pid-training-process-dynamics 16/27
Time
100%
63.2%
0%
High-order system
Approximated to first order
deadtime plus lag
% of finalvalue
DERIVING PROCESSDYNAMICS
The dynamics may be derived in two ways
The first is to approximate the curve to one of deadtime and 1st order lag and to draw a tangent at the point of steepest slope. The point of intersection on the X-axis marks
the end of the deadtime and start of the lag period
8/16/2019 02 - PID Training - Process Dynamics
http://slidepdf.com/reader/full/02-pid-training-process-dynamics 17/27
Time
100%
0%
20%
80%
T20 T80
Deadtime = t20 - 0.161 (t80-t20)
Lag = 0.721 (t80-t20)
Steady-state value
DERIVING PROCESS DYNAMICSWITH THE 20%/80% METHOD
Another and more easy method is the 80/20 method developed by Honeywell Hi-Spec.
It uses a simple empirical formula to calculate the dynamics directly.
Simply measure 20% and 80% of the final value on the Y-axis and then determine the corresponding times. Then use theformula above
8/16/2019 02 - PID Training - Process Dynamics
http://slidepdf.com/reader/full/02-pid-training-process-dynamics 18/27
EFFECT OF VARYING
DYNAMIC PARAMETERS
Increasingdeadtime
Increasing Gain
IncreasingTime Constant
IncreasingOrder
Order = 0
! = 0
8/16/2019 02 - PID Training - Process Dynamics
http://slidepdf.com/reader/full/02-pid-training-process-dynamics 19/27
INVERSE RESPONSE (1)
PC
8/16/2019 02 - PID Training - Process Dynamics
http://slidepdf.com/reader/full/02-pid-training-process-dynamics 20/27
INVERSE RESPONSE (2)
Time
Time
Valve
Position
Pressure
Inverse Response
8/16/2019 02 - PID Training - Process Dynamics
http://slidepdf.com/reader/full/02-pid-training-process-dynamics 21/27
STABLE / INTEGRATINGPROCESS
Ti
Fuel gas
Self-regulating Process Non-Self Regulating Process
If increase opening of fuel gas control valve,
then coil outlet temperature will rise and move
to a new steady-state value
If increase opening of outlet valve, then
level will fall and continue to fall without
reaching a new steady-state value
LI
FC
8/16/2019 02 - PID Training - Process Dynamics
http://slidepdf.com/reader/full/02-pid-training-process-dynamics 22/27
LINEARITY
• A linear system has a constant process gain
• Most processes in refining, chemicals etc are atleast slightly non-linear
• This has implications for controller tuning(need to tune for different conditions)
• Implications for plant tests or step tests
• Some processes are known to be highly non-linear e.g. pH control
8/16/2019 02 - PID Training - Process Dynamics
http://slidepdf.com/reader/full/02-pid-training-process-dynamics 23/27
LINEAR / NON LINEAR
Process Variable, PV
Manipulated
Variable, MV
Linear
Highly non-lineare.g. pH
pH = -log10 [H+]
(Note a system is linear if
dPV/dMV is constant)
8/16/2019 02 - PID Training - Process Dynamics
http://slidepdf.com/reader/full/02-pid-training-process-dynamics 24/27
PLANT TESTS,GUIDELINES 1
• Choose the right time
- Not at a shift changeover
- Not during a feed switch
- Not at dawn or dusk
- Not on Monday morning!
- Not on Friday afternoon!
• Set up sampling/lab analysis if required
• Talk about test with operations and get approval for test
• Ensure instrumentation is OK
• Ensure process is steady
Open loop response is required for getting accurate tuning parameters. With out step test, is it possible deriving open loop response from closed loop data?Yes...explain it through bock diagram.
8/16/2019 02 - PID Training - Process Dynamics
http://slidepdf.com/reader/full/02-pid-training-process-dynamics 25/27
PLANT TESTS,GUIDELINES 2
• Open required loop, make step change of say 2-5% valvemovement
• Record trend of OP and PV
• Reach new steady state some people use !+5"• Carry out test again in opposite direction, double amount if
possible
• Ensure Manipulated variable really changes
• Document results
• Do tests under all modes of operations
• Step away from any constraints on the unit for safety
8/16/2019 02 - PID Training - Process Dynamics
http://slidepdf.com/reader/full/02-pid-training-process-dynamics 26/27
TYPICAL PLANT TEST
Time
MV CV
Time
8/16/2019 02 - PID Training - Process Dynamics
http://slidepdf.com/reader/full/02-pid-training-process-dynamics 27/27
CONCLUSION
• Understanding dynamics is the key to
controlling any process
• Simple dynamics can be represented by
gain, deadtime and lag
• These may be determined by carrying out
plant step tests