Level Control of a Wood Chips Tank
Snapshot of the front panel of the simulator:
The level control system for a wood chips
tank is simulated. The purpose of the level control is to keep the level
between specified limits: It is important that the level is not too high,
otherwise the chip will not be sufficiently preheated (by steam from the
cookery), or the tank will simply run full. And it is important that the
level is not too low, otherwise the steam will stream through the chip
causing an awful smell in the neighbourhood.
The level control system is inspired by, and is similar to, the level
control system at Tofte SÃ¸dra Cell plant in Norway. (The physical plant does not exist any more, but the simulated plant is still alive (here).)
The process consists of a chip tank with an inlet screw,
a conveyor, and an outlet screw. The inlet flow can be controlled by
adjusting the screw control variable. The outlet flow can not be
controlled, and it constitutes a disturbance on the level.
Mathematical description of chip tank, level sensor and measurement
filter
Below is a mathematical model of the process. However, knowledge about
this model is not necessary to do the tasks below.
Process model
A mathematical model of the process can be found using a
mass balance:

A*r*dh(t)/dt
= u(tt_{transp})  F_{out}(t) 
(Eq. 1) 
 h [m] is the level. h is the process output variable,
which is to be controlled.
 u [kg/s] is the control variable, acting on the inlet screw, and giving
a proportional mass flow through the screw.
 A [m^{2}] is the cross sectional area of the
tank.
 r [kg/m^{3}]
is the chip density.
 t_{transp} [s] is the transport
time of the conveyor belt, which runs with constant speed. t_{transp}
can also be denoted as dead time or time delay.
 F_{out} [kg/s] is the mass outflow from the
tank. F_{out} is a disturbance on the level.
Dynamically, this process is "an integrator with deadtime"
from the screw control.
The values of the model parameters are available via the front panel of the
simulator.
Level measurement and measurement filter
The level is measured with an ultrasound level sensor.
Measurement random noise is added to the level measurement. The
measurement signal passes through a lowpass filter which attenuates the
noise component of the signal. The filter is a moving average filter with filter time constant t_f [s].
The filter action is removed by setting t_f = 0.
The aims of the tasks given below
are
 to give an understanding of how an automatic feedback
control system works, and which benefits feedback control has compared
to using a fixed value of the manipulated variable
 to give an understanding of the properties of PID control
 to develop skills in tuning controller parameters
 to give insight into how various parameters (in the controller, in the
process, and in the measurement device) influates
the dynamic properties and the stability properties of the control
system.
In other words: This lab will give your (more) insight
into the most important issues of control engineering!
Control systems are essential in industry because it is
important to control process variables so that they are kept equal to or
close to setpoints. The PIDcontroller is by far the most frequently used
controller function, and it is a main topic in this lab.
This lab is about level control, and in most plants there is a need to
control level. The simulated tank with it's level control system in this
lab is a "real" system, as it actually exists, as mentioned
above.
In the tasks below it is assumed that the process is in
it's nominal operating point unless something else is stated. The
nominal operating point is defined as follows:
 The level reference (setpoint) is 10 m.
 The chip outflow is F_{out} = 25
kg/s = F_{out,nom}.
 The nominal value of the manipulated variable is u_{nom} =
45%, which gives an inflow that is equal to the nominal outflow.
 The moving average filter time window is t_ma = 20 s.
In the tasks about PIDcontrol: Set the setpoint
weights w_{p} and w_{d} equal to 1, and the coefficient a
= Tf/Td equal to 0.1 (these are also the default values as set on the
front panel). Let the PIDcontroller have antiwindup.
 First: No controller!
Set the controller in manual mode. Give the outflow a step from e.g. 1500
to 1800 kg/min, which implies that u_{nom} no longer fits to
the nominal outflow, F_{out,nom}. Characterize the response in
the level. Is control using a fixed value of the manipulated variable
(u_{nom} = constant) an acceptable way of
controlling this process?
 Then: Manual control,
that is: YOU are the controller! Set the controller
in manual mode. Give the outflow a step from e.g. 25 to 30 kg/s.
Compensate for this disturbance by adjusting u_{nom}. How long
time do you need to bring the level back to the setpoint with an
error less than 0.1 m? Are there any drawbacks with using a human
being as a continuous controller?
 Automatic control with
a PIDcontroller: Parameter tuning
Tune the PIDcontroller using
the ZieglerNichols' closedloop method.
In the following tasks you shall use the
PIDparameter values as found in task 4. Alternatively, you can use the following PIDparameter values,
which you can regard as "standard values":
Kc = 3.89, Ti = 1000 s, Td = 0 s =
135 sec.
 Is the stability of the
control system OK? Give w_{ut} a step
from e.g. 1500 to 2000 kg/min, and observer how the level returns to the
setpoint. Does the control system have proper stability?
 Compensation properties:
 How large is the stationary control error using a
PIDcontroller after a step in w_{ut} from 1500 to 1800 kg/min?
 How long time time does it take for the
PIDcontroller to bring the level back to the setpoint with an
error smaller than 0,1m (after the step in w_{ut} as
described above)?
 Use a Pcontroller. Choose a proper value of Kc
in the Pcontroller. How large is now the stationary control error?
Zero?
 Tracking properties:
How large is the stationary control error with a PIDontroller after a
step in the setpoint (choose the step value yourself)?
 How the P, I and Dterm woks:
Observe how the three terms in the control signal, u, works after a step
in the disturbance. (There is a button beneath the diagram of u to show
the timeresponse of the individual control signal terms.)
 How parameter changes
influences the stability of the control system: Observe
how the stability of the control system changes due to the parameter
changes describes below. In each subtask/experiment you can excite the
control system with a small step in the setpoint. The experiments
must be performed independant of each other, that is, you have to
reset the parameters to the standard values (defines above) between each experiment.
 The controller gain K_{p} is increased (much).
 The integral time T_{i} is reduced (much).
 The derivate time T_{d} is increased (much).
 The transport time (time delay) t_{transp} of the conveyor is increased (much).
 The cross sectional area A is reduced (much).

The importance of using
the correct measurement function in the setpoint generation:
 What happens with the control error of the
setpoint gain K_{m,LC} is different from the measurement
gain K_{m,LT}?
 What happens with the control error if the
lower limit, r_{m0}, of the setpoint range is different
from the lower limit, y_{m0}, of the measurement range?

Measurement lowpass filter:
Demonstrate that the measurement lowpass filter attenuates the noise
component of the measurement signal. Also demonstrate that the filter
reduces the variation of the control signal. (Compare the behaviour of
the signals with filter and without filter.) In general,
what may be the drawback of large variations of the control signal?

Derivative term and measurement
noise: Demonstrate that the derivative term of the controller
causes the control signal to vary more (compared to not having
derivative term).
[SimView] [TechTeach]
Updated 7th December 2023.
Developed by
Finn Haugen.
Email: finn@techteach.no. 