Improve Control
of Liquid
Level Loops
Robert Rice Use this tuning recipe
Douglas J. Cooper
for the classic integrating
B
Control Station, Inc.
process control challenge.
ecause most processes are self-regulating, it can is increased by a fixed amount, the car will accelerate and
sometimes be challenging to tune a controller for an then settle at a different constant speed.
integrating process. The principal characteristic of a The temperature of a process stream exiting a heat
self-regulating process is that it naturally seeks a steady- exchanger is also self-regulating. If the shellside cooling
state operating level if the controller output and disturbance fluid flowrate is held constant and there are no significant
variables are held constant for a sufficient period of time. external disruptions, the tubeside exit stream temperature
For example, a car’s cruise control is self-regulating. will settle at a constant value. If the cooling flowrate is
By holding the fuel flow to the engine constant (assuming increased, allowed to settle, and then returned it to its
the car is traveling on flat ground on a windless day), the original value, the tubeside exit stream temperature will
car is maintained at a constant speed. If the fuel flowrate move to a new operating level during the increased
flowrate and then return to its original steady-state.
Tanks that have a regulated exit flow stream do not nat-
Self-Regulating urally settle at a steady-state operating level. This is a
common example of what process control practitioners
PV
Beer
Surge tanks are often installed between two Storage
process systems with incompatible flow patterns Beer Pump
to provide flow smoothing. The “wild stream” has
FT02
flow control requirements that are difficult to influ- CT01 FLOW
CO2
ence, and the controller then adjusts the controlled
FT01
Beer CT04
Tank
TT02 AT02 PT03
Upper Constraint
are viewed as poor performers and are then 60
tuned for tight performance, counteracting the
intended design objective. 50
A major beer brewer uses an SVK system to
fill several lanes of kegs (top). Because the keg- Lower Constraint
40
filling lanes are operated in an on/off fashion, the
wild stream flowrates requested by the SVK sys- Aggressively Tuned PI Controller Conservatively Tuned PI Controller
120
ing the flow of beer pumped from the large stor- 100
age tanks controls the level in the surge tank. 80
Due to the sensitive nature of beer and of the 60
40
analytical instrumentation involved, a surge tank 20
is installed to dampen the large demand fluctua- 0
tions required by the keg-filling system. By allow- 0 5 10 15 20 25 30
The integrating behavior plot is somewhat misleading, as shown in Figure 2 (p. 56), when the setpoint (SP) is ini-
it implies that for such processes, a steady controller output tially at the design level of operation (DLO) in the first
will produce a steady process variable. While this is possi- moments of operation, then PV equals SP (the DLO is
ble with idealized simulations like that used to generate the where the setpoint and process variable are expected to be
plot, such “balance point” behavior is rarely found in inte- during normal operation when the major disturbances are
grating processes in industrial operations. at their normal or typical values).
More realistically, if left uncontrolled, the lack of a bal- The setpoint is then stepped up from the DLO on the
ance point means that the process variable of an integrating left side of the plot. The simple P-only controller is unable
process will naturally tend to drift up or down, possibly to to track the changing SP, and a steady error, called offset,
extreme and even dangerous levels. Consequently, integrat- results. The offset grows as each step moves the SP farther
ing processes are rarely operated in manual mode for long. away from the DLO.
Midway through the process, a disturbance occurs, as
P-only control behavior is different shown in the middle of the plot. (Its size was predeter-
To appreciate the difference in controlled behavior for mined for this simulation to eliminate the offset.) When
integrating processes, first consider the proportional, or the SP is then stepped back down (on the right) the offset
P-only, control of an ideal self-regulating simulation. As shifts, but again grows in a similar and predictable pattern.
65
60
P-only control, as shown on the left of Figure 3, the process
variable is able to track the setpoint steps with no offset. This
55
50
controller does not need to.
Yet the setpoint steps in the right of Figure 3 show this is
not completely correct. Once a disturbance shifts the baseline
100 200 300 400 500 600
Time
or balance-point operation of the process (shown roughly at
the midpoint in the plot), an offset develops and remains con-
■ Figure 2. P-only control of an ideal self-regulating process
65
52
40 80 120 160
Time PI control behavior is different
The dependent, ideal form of a proportional-integral (PI)
■ Figure 3. Unlike an ideal self-regulating process, P-only control controller (1) is one of numerous algorithms that are widely
employed in industrial practice:
for an ideal integrating process shifts the baseline operation of the
K
process, producing a sustained offset even as the setpoint returns
56 57
PV and SP, %
54 54
52 51
50 48
48
45
Kc = 0.3 Kc = 0.3 Kc = 1.2 Kc = 1 Kc = 4 Kc = 8
60 No oscillation Modest oscillation PV oscillates PV oscillates No oscillation PV oscillates
100
80
CO, %
CO, %
55
60
50
40
45 20
2. Bump the process, and collect dynamic process data of be computed using a graphical analysis of plot data, or in an
the process variable response to changes in controller output. industrial setting by automated analysis using a commercial
3. Approximate the process data behavior with a first- software package. Once the model parameters are known, the
order-plus-dead-time integrating (FOPDT integrating) tuning values for the dependent, ideal PI form, Eq. 1, as well
dynamic model. as the popular PID algorithm form, can be calculated:
4. Use the model parameters generated in step 3 and the K dPV
correlations in Table 1 to complete the controller tuning. CO = CObias + K c e(t ) + c ∫ e(t )dt − K c Td (5)
Ti dt
It is important to recognize that real processes are
more complex than the simple FOPDT integrating model. For integrating processes there is no identifiable
In spite of this, the model does provide an approximation process time constant in the FOPDT integrating model.
of process behavior that is sufficiently rich in dynamic Thus, dead time, θp, is used as the baseline marker of time
Level, m
2
0
Exit flow increases …
80
CO, %
75
70
20 25 30 35 40 45
Time, min
the total flow into the tank is less than the flow pumped
out, the liquid level will fall and continue to fall.
Figure 7 is a plot of the pumped-tank behavior with the
controller in manual mode (open-loop). The CO signal is
■ Figure 6. Simulated pumped-tank level control in automatic stepped up, increasing the discharge flowrate out of the
mode uses a throttling valve to adjust the process variable, the
bottom of the tank. The flow out becomes larger than the
total feed into the top of the tank and, as shown, the liquid
liquid level in the tank.
for tuning. Specifically, θp is used as the basis for comput- level begins to fall. As the situation persists, the liquid
ing the closed-loop time constant, T c. level continues to fall until the tank is drained. The saw-
Building on the popular internal model control (IMC) toothed pattern occurs when the tank is empty because the
approach to controller tuning, the closed-loop time con- pump briefly surges every time enough liquid accumulates
stant is computed as T c = 3θp (3). for it to regain suction.
The controller tuning correlations for integrating Figure 7 does not show that if the controller output were
processes use this T c, as well as the K p* and θp from the to be decreased enough to cause the flowrate out to be less
FOPDT integrating model fit, in the correlations of Table 1. than the flowrate in, the liquid level would rise until the
tank was full. If this were a real process, the tank would
A simulated example — the pumped-tank overflow and spill, creating safety and profitability issues.
A pumped-tank simulation illustrates the design and tun-
ing of a controller for an integrating process. As shown in Graphical modeling of integrating process data
Figure 6, the process has two liquid streams feeding the top The graphical method of fitting an FOPDT integrating
of the tank and a single exit stream pumped out of the bot- model to process data requires a data set that includes at
tom. The measured process variable (PV) is the liquid level least two constant values of controller output, CO1 and
in the tank. To maintain the liquid level, the controller out- CO2. As shown in Figure 8 for the pumped tank, both must
put (CO) signal adjusts a throttling valve at the discharge of be held constant long enough that a slope trend in the PV
a constant-pressure pump to manipulate the flowrate out of response (tank liquid level) can be visually identified.
the bottom of the tank. This approximates the behavior of a An important difference between the traditional process
centrifugal pump operating at relatively low throughput. reaction curve graphical technique for self-regulating
Note that a pump strictly regulates the discharge processes and integrating processes is that integrating
flowrate out of the tank. As a consequence, the physics do processes need not start from a steady-state value before a
not naturally work to balance the system when any of the bump is made to the CO. The graphical technique discussed
stream flowrates change. This lack of a natural balancing here is only concerned with the slopes (or rates of change)
behavior is why the pumped tank is classified as an inte- in PV and the constant controller output signal that caused
grating process. If the total flow into the tank is more than each PV slope.
the flow pumped out, the liquid level will rise and contin- The FOPDT integrating model describes the PV behavior
ue to rise until the tank fills or a stream flow changes. If at each value of constant controller output, CO1 and CO2, as:
Level, m
4.4 4.4 (36, 4.6)
Slope1 (24, 4.8)
4.0 4.0
75 75
CO, %
CO, %
CO1
70 70 CO1 = 65 CO2 = 75
CO2
65 65
20 25 30 35 40 20 25 30 35 40
Time, min Time, min
■ Figure 8. To perform a manual-mode bump test of the pumped- ■ Figure 9. The slopes are calculated from bump test data to
tank process, the controller outputs must be held constant long compute the integrator gain, Kp*.
enough to show the slope trend in the PV response.
dPV
= K *p CO1 (t − θ p )
dt 1 5.2
and 4.8
Level, m
dPV
= K *pCO2 (t − θ p ) (6))
4.4
dt θP = 1 min
4.0
2
dPV dPV
CO, %
− 70
dt dt 1 Slope2 − Slope1
K *p = 2 = (7)
CO2 − CO1 CO2 − CO1
65
20 25 30 35 40
Time, min
Computing integrator gain. The values of the open-loop
Graphical modeling of pumped-tank data
troller output is then stepped from 65% up to 75%, causing is normally put into practice using bumpless transfer —
a downward slope in the liquid level. that is, when switching to automatic control, SP is initial-
The slope of each segment is calculated as the change in ized to the current value of PV and CObias to the current
tank liquid level divided by the change in time. From the value of CO. By choosing the current operation as the
plot data, Slope1 is calculated to be 0.13 m/min and Slope2 design state at switchover, the controller needs no correc-
as –0.12 m/min. Using the slopes with their respective CO tive actions and it can smoothly engage.
values yields the integrator gain, K p* = –0.025 m/%-min. Controller gain, K c, and reset time, T i. The first step in
Computing dead time. The dead time, θp, is calculated as using the IMC correlations listed in Table 1 is to compute
the difference in time from when the CO signal was stepped T c, the closed-loop time constant. T c describes how active
and when the measured PV starts to exhibit a clear response the controller should be in responding to a setpoint change
to that change. From the plot in Figure 10, the pumped-tank or in rejecting a disturbance. For integrating processes, the
dead time is estimated be θp = 1.0 min. design and tuning recipe suggests:
D, %
the top of the column. 3
Distillation columns are very sensitive unit operations 2
with very slow response times (long time constants). If the 1
level controller is tuned aggressively for tight setpoint track- 10 20 30 40 50 60 70 80
Time
ing, large and rapid reflux flow changes could dramatically Sample Time, T = 1 s
impact column efficiency and stability. Thus, the reflux drum
needs to be tuned for conservative control actions while ■ Figure 11. A PI controller provides setpoint tracking and
maintaining the level constraints. disturbance rejection.
Using the tuning procedure outlined in this article, the
1 2Tc + θ p
reflux drum level not only tracks closer to setpoint — it does
Kc = ⋅ Ti = 2Tc + θ p (8)
K *p (Tc + θ p )2
so with 95% less controller output movement.
60 Upper Constraint
Literature Cited
90
30 Aggressively Tuned PI Controller Conservatively Tuned PI Controller
80
Reflux Flow, %
K p* = integrator gain
workshops. Prior to joining Control Station, he was an engineer with PPG
Industries. He received his BS in chemical engineering from Virginia
SP = setpoint
engineering from the Univ. of Connecticut.
T = sample time
Station, Inc. (Phone: (860) 872-2920; E-mail:
θp
intelligent technologies and adaptive process control; and software tools
for process control system analysis, tuning and training. Prior to forming
Control Station, he held research positions with Arthur D. Little and
Chevron. He received his BS in chemical engineering from the Univ. of
Massachusetts, Amherst, MS in chemical engineering from the Univ. of
Michigan, and PhD in chemical engineering from the Univ. of Colorado.