Figure 1 self regulating process
About Tune Guide
Tune Guide is a program witch will assist you in selecting the correct PI parameters for industrial controllers. The program calculates gain and integration time based on three parameters to describe the process, this means that only first order processes are suitable to tune with this program. At the other hand the majority of loops in the process industry are just of first order. The Amigo method are suitable even for slightly more complicated processes. But for complicated processes other tuning methods like Dominant Pole Placement are recommended. the parameters Kp (Process gain), T (Time constant) and L (Dead time) are easily determined by doing a step response experiment, how to perform the experiment are described below. The controllers sampling time Ts, is used for calculation of the transmitters filter time-constant. For optimal filter setting the filter time should be reduced to 1,3*sample-time if possible, when the experiment are performed. If the filter time due to noise had to be increased a second step response experiment have to be done.
Not ALL processes can be tuned with one and the same method, different processes require different tuning methods. Lambda tune give quite safe parameters and the possibility to select the speed of the closed loop, it also minimizes actuator wear, but in some cases the response on load rejections can be extremely slow. The Amigo method on the other hand are very powerful and give parameters within the same range as some commercial tuning tools. This method maximizes integral gain witch is the key contribution to load rejection.
Please note: that set point weighting should be used to avoid undesired overshots from set point changes when the Amigo methods are selected, because this method give parameters suited for load rejection. The beta can be set to 0,5-0,8 as a starting value, fine tuning of beta by hand are recommended. If your controller don’t support set point weighting it’s recommended to limit the speed of a set point change.
When this was written, only the ABB controller PIDCONA supported variable Beta
values. For KLT processes (a pure first order process). Amigo KLT is developed, it is a modified calculation especially adapted for pure first order (or monotone) processes. If Amigo do not give the requested load rejection, then you can try the Cohen-Coon or Ziegler-Nichols methods, those methods are more aggressive but remember that the design criteria for the last two is quarter amplitude damping. It’s very likely that a controller tuned in that manner goes into oscillation if the process parameters Kp, T or L changes. The units used in this program are for P=Gain and the Integration time is=sec/rep.
How to determine Kp, T and L? (for self regulating processes)
The easiest way to determine the process parameters are to perform a step response experiment. You will need a logger or a fast printer in order to record the process variable and the controllers output, for a flow loop 100-200ms scan rate are recommended. The DCS screens are often not fast
enough.
1. First put the controller in manual mode and wait for steady state
2. Make a small output change so the Process variable reacts, this is done to overcome eventual
hysterias.
3. Do a bigger output change in one step (the red line in the picture below)
4. Wait for the process variable to reach a new steady state
5. Return the controller to Auto
6. Calculate the process parameters as follows:
Kp = Delta PV / Delta OUT
T = The time needed for the PV (Process variable) to reach 63% of Delta PV, measured from the point were the PV started to move. Expressed in seconds.
L = The time from the output change to the point were PV starts to move. (seconds)
The parameters for the step response below are approximate
Kp=0.96 T=3.6 sec. L=0.7 sec.
Figure 1 self regulating process
Program handling:
1. Fill in the process parameters Kp, T, L and Ts.
2. Select tuning method in the tuning method frame.
3. The proposed parameters becomes visible at the right
4. The suggested filter time and the controllers scan time are presented in the “Settings”
frame.
5. Change tuning method and compare the results
6. Use the Calculate button if you have changed one or several process parameters and wish to
use the last selected tuning method. Or if you use the time-constant function under Lambda
tune.
7. Save the loop for later use, supply your own comments in the notes field.
8. A tool tip text is available if you point with the mouse at any text label.
Notes: When using Lambda tune as tuning method and the process are dead time based i.e. T < (L + Ts) the program forces the lambda to (3 * T) or (3 * L + Ts) depending on witch combination having the highest value. For enhanced control select Time constant mode and enter a lower time constant, the program will display a error message if the selected time constant are too low.
The gain proposed by AMIGO and AMIGO KLT may be high for some processes, in this case the gain could be reduced and the integral time can be left unchanged. However the design parameter Ms, is selected for a non-oscillating behavior. For the most processes the gain can actually be doubled if an overshot from a load change are acceptable.
Integrating processes
The program needs the Velocity gain (Kv) witch is how much the process variable can move per second, and the dead time (L). As for the stable process a simple step response experiment is enough for determine the process parameters Kv and L.
1. Put the controller in manual mode and wait for steady state or constant increasing or
decreasing process variable
2. When the trend of the PV are clear, do a step on the controllers output
3. Wait for a reaction on the PV, allow the PV to move for some time
Calculate Kv and L as follows
Kv = Delta PV / Delta OUT / Delta Time (in seconds)
L = The time from the output change to the point were PV reacts expressed in
seconds.
When using Lambda tune the desired set point and expected load change will be used as design criteria’s. 50% set point and 50% load change are selected as default. Press the Calculate button after a change of values.
NOTE: It’s almost newer needed to select higher load change then 50%, though the gain may be unnecessarily high
1. Fill in the process parameters. Kv, L and Ts.
2. Select tuning method in the tuning method frame. Amigo or Lambda tune are
implemented.
3. The proposed parameters becomes visible at the right
4. The suggested filter time and the controllers scan time are presented in the “Settings”
frame.
5. Change tuning method and compare the results
6. Use the Calculate button if you have changed one or several process parameters and wish to
use the last selected tuning method. Or if you use the load change / SP functions under Lambda
tune.
7. Save the loop for later use, supply own comments in the notes field.
The picture below shows a step response experiment.
Were Kv = 0,180952
And L = 23 seconds.
Figure 2 integrating process
Testing the valve for sticktion:
1. Put the controller in manual mode and wait for steady state.
2. Make a 1% step in the controller output (green curve in figure 3)
3. Study the process variable (Blue curve), if no reaction is seen, make another 1% step in the same direction.
As seen in the graph below the controller output needs to move 2% before the PV reacts. The stiction will make this loop very difficult to tune, and it will create a oscillation around the set point. The effect of the loop can as a general rule be multiplied with the process gain, the Kp value. If the Kp are 2,3 the effect on the loop will be 2.3 * 2 % = 4.6 % witch is totally unacceptable. This valve should be maintained as soon as
possible.
Figure 3 sticktion
Testing the valve for hysterias and sticktion:
1. Do the same test as for the sticktion test above
2. When the PV reacts do one or two more steps in the same direction.
3. If the PV reacts on the additional output changes, the valve do not have a sticktion problem.
4. Change direction and continue with 1% changes until PV reacts.
As seen in the graph below the controller output needs to move 2% before the PV reacts. But after that the PV reacts for every 1% change. The phenomena repeats itself after the change of direction of the manual output changes. The hysterias origin is mostly wear in the actuator or the valve shaft. The hysterias will make this loop very difficult to tune, and it will also create a oscillation around the set point. But the period time will be longer. This valve should be maintained as soon as possible.
Figure 4 hysterias