THE COPPERBELT UNIVERSITY

School of mines and minerals sciences

CE 560 / MT 580
Process control

Lecture 09
Integral, derivative and PID controls

C. Botha (Mr.)
Contacts: Chemical Engineering Department
Email: [email protected]
Alt. Email: [email protected]

January, 2024
 Integral (I) control

o Proportional control fails to eliminate offset in a response.

 This is the main reason why other control actions such as integral are preferred.
o Integral control is widely used because it provides an important practical advantage.
 That is, offset (steady state error) is completely eliminated from the response,
 With integral control, controller output is varied proportionally to the time
integral of the error.
 Hence, integral control law can be expressed as
1 𝑡
𝐶𝑜 = 𝐶𝑂𝑏𝑖𝑎𝑠 + ‫𝜀 ׬‬ 𝑡 𝑑𝑡 9.1
𝜏𝐼 0
 Integral (I) control

 The new adjustable parameter 𝝉𝑰 is known as integral time or reset time.

 Eq. 9.1 implies that the integral term integrates or sums-up the instantaneous
errors.
 That is, sums up the difference between SP and PV over the duration for which
this error exits.
o The time-integral error amount is then added or subtracted from the nominal
controller output to give the final controller output signal.
o Integral control law (Eq. 8.1) is typically written in the form
𝑡
𝐾𝐶
𝐶𝑜 = 𝐶𝑂𝑏𝑖𝑎𝑠 + න 𝜀 𝑡 𝑑𝑡 9.2
𝜏𝐼
0
 Integral (I) control

o Where
𝜀 𝑡 = 𝑦𝑠𝑝 − 𝑦𝑚 𝑡 𝑜𝑟 𝜀 𝑡 = 𝑆𝑃 − 𝑃𝑉 9.3

o Eq. 9.2 is the form conventionally used in commercial controllers.

o To understand why integral control eliminates the steady-state error,

 Consider the integral control law, Eq. 9.2.

 Controlled variable is at steady state if the controller output ( and hence the
manipulated variable) is constant.

 Eq. 9.2 implies that CO continuously changes until 𝜀 𝑡 = 0.

 Thus, with integral control, controller output is continuously adjusted.

 Integral (I) control

 Likewise, the manipulated variable is adjusted as a result of the changing controller

output.

 Adjustment only stops when the controller output reaches a value required to
make the error zero.

 That is, controlled variable becomes equal to the desired value.

o For the STH with 𝑲𝑪 set to its highest value.

 The controlled variable will attain its new equilibrium value with P-control.

 However, when integral action is added, it will continue to sum-up the error, no
matter how small.
 Integral (I) control

 Thus, controller output and hence steam flow are continuously adjusted.
 Adjustments continue until controller output attains a value required to make
the steady state error or offset zero.
 That is, when exit temperature reaches desired value of the temperature.
 This desirable situation always occurs with integral action unless the final control
element or controller saturates.
Proportional plus integral control
o Despite eliminating offset, integral control is rarely used by itself.
o Consider the system’s response shown in Fig 9.1.
 Integral (I) control

o The set point is changed according to a

step increase.
o Process variable will track-up the change
in the set point.
o Response may overshoot if 𝐾𝐶 is too
large.
o Since P-control action will be aggressive.
Fig. 9.1 PI-controller response to a set-point
o From Fig. 9.1 and P-control law, it can be change
seen that the proportional term has
immediate effect on controller output.
 Integral (I) control

o This results in immediate corrective action as soon as the error is present.

o I.e. immediately process variable changes or in case of Fig. 9.1 set point changes.

o Integral term on the other hand has to integrate the error.

o Hence, little integral control action takes place until the error exists for some time.

o Thus integral action takes some time before having a significant impact on the
process response.

o This is because the integral (summation) of the error must be significant before
integral control can produce a reasonable corrective action.
 Proportional plus integral control

o Since P-control takes immediate corrective action i.e., as soon as the error is detected.
o Integral control is typically used in conjunction with proportional control.
o This forms proportional plus Integral (PI) control for which the control law is
𝑡
𝐾𝐶
𝐶𝑂 = 𝐶𝑂𝑏𝑖𝑎𝑠 + 𝐾𝐶 𝜀 𝑡 + න 𝜀 𝑡 𝑑𝑡 9.4
𝜏𝐼
0
𝑡
1
𝐶𝑂 = 𝐶𝑂𝑏𝑖𝑎𝑠 + 𝐾𝐶 𝜀 𝑡 + න 𝜀 𝑡 𝑑𝑡 9.5
𝜏𝐼
0

o That is controller output depends on the error itself and the time integral of the error.
 Proportional plus integral control

o The integral time is an adjustable parameter and 𝝉𝑰 (in minutes) is referred to as

minutes per repeat.
o Value of the integral time is usually in the range 0.1 ≤ 𝜏𝐼 ≥ 50 𝑚𝑖𝑛.
o Some controllers are calibrated in 𝝉𝑰 whiles others are in terms of 𝟏Τ𝝉𝑰 referred to
as repeats per minute.
o The repeats per minute is known as reset time for the reason as follows:
o Suppose the error changes by a step of magnitude 𝜀 and stays there for a time 𝝉𝑰 .
o From equation 9.2, we get:
𝑡
𝐾𝐶 𝐾𝐶
𝐶𝑂 = 𝐶𝑂𝑏𝑖𝑎𝑠 + න 𝜀 𝑡 𝑑𝑡 = 𝐶𝑂𝑏𝑖𝑎𝑠 + 𝜀𝜏𝐼 = 𝐶𝑂𝑏𝑖𝑎𝑠 + 𝐾𝐶 𝜀 9.6
𝜏𝐼 𝜏𝐼
0
 Proportional plus integral control

o Thus, integral action makes the same contribution to the controller output as
proportional control.
o Repetition takes place every 𝝉𝑰 minutes hence the name repeats per minute or
reset time.
Drawbacks of I or PI control
o One limitation of integral control is the lack of immediate response to disturbances.
o Other limitation arises from the dependence of integral action on the time integral of
the error.
 Proportional plus integral control - drawbacks

o A sustained error results in large integral term leading to a large +ve or –ve controller
output.
o Such a controller output may be meaningless e.g., may require valve opening of 150%
or -20%.
o That is values that are outside the final control element range, a condition known as
controller saturation.
o Further build up of the integral term while the controller is saturated is known as reset
wind-up or integral wind-up.
o Repetition of integral action ceases when this occurs.
 Proportional plus integral control - drawbacks

o Integral wind-up occurs when a PI and PID controller experiences a sustained error.

o E.g., when starting-up of a unit, or after a large set-point change or a large sustained
disturbance occurs.

o Commercial controllers have anti-reset wind up build in their algorithms.

o This limits the contribution of the integral term after controller saturation.

o Integral control lacks immediate corrective action.

o In addition, it results in undesirable effects due to accumulation of integral term.

o I.e., tends to produce oscillatory response.

 Proportional plus integral control - drawbacks

o Since build-up of integral term can lead to a sensitive or aggressive response.

o This may produce oscillatory response, especially where the error persists.

o Proper tuning or including derivative action can reduce the effects of integral term.
Tuning guidelines
o Fig. 9.1 depicts the dependence of the response on the tuning parameters 𝐾𝐶 and
𝜏 𝐼 for a PI-controller.

o Controller tuning implies arriving at the best combination of the tuning parameters
that result in the desirable response.
o 𝐾𝐶 and 𝜏 𝐼 are interacting and makes tuning a PI controller challenging.
 PI - control – tuning guidelines

2Kc

0 Time 0 Time Time

Kc

Time Time Time

Kc /2

Time Time Time

τI / 2 τI 2τI
Increase

Fig.9.2 Sensitivity of process response to Kc and τI

 PI - control – tuning guidelines

o Increasing 𝑲 𝒄 while 𝝉 𝑰 is kept constant results in increase in both integral and

proportional terms.
o This, as can be seen from Fig. 9.2, makes the response faster or more
aggressive (sensitive).
o Unfortunately, aggressive response has a tendency to produce oscillations in the
response.
o Increasing the value 𝝉 𝑰 while 𝑲 𝒄 is kept constant decreases the integral term.
o This may results in a less aggressive or slow response and less oscillations.
o Hence, proper controller tuning will seek to find the best combination of 𝑲 𝒄 and 𝝉𝑰.
 Derivative control

o In derivative (D) control, the controller output varies proportionally to the rate of change
of the error, therefore;
𝑑𝜀 𝑡
𝐶𝑂 = 𝐶𝑂𝑏𝑖𝑎𝑠 + 𝐾𝐶 9.7
𝑑𝑡

o Where 𝝉𝑫 is the derivative time, in units of time and is another tuning parameter.
o Derivative control depends on the rate of change of error.
o I.e., considers current rate of change of the error to anticipate future behaviour of the
error.
 Control action was referred to as rate action or anticipatory control.
 It contributes to controller output as soon as there is a change in error.
 Derivative control

o Its action is similar to P-control with high gain for rapidly changing measured variable
and low gain for a change that is not as rapid.

o Consider an increase in furnace temperature; 10℃, say, in 1 min, and 30℃ in 30 min.
 Increase in temperature in the first case is more rapid than in the second case.
o Manual control, operator observes how quickly the temperature is rising.
 Anticipates how large the deviation in temperature will be and its consequences.
 Appropriate control action is then taken by the operator to correct the temperature.
 Action depends on how quickly the temperature is rising and the anticipated
deviation.
 Derivative control

o Derivate control works in a similar manner, that is,

 Anticipates future behaviour of the error and its effect, then take appropriate action.
o Proportion and integral control would not be effective for such a sudden change in error.
o Since P-control depends on magnitude of the error and not how fast the error changes.
o Integral control depends on the period for which error exists and not how rapid the error
changes.
o Thus, would not be effective for a sudden deviation in temperature.
o Derivative control does not eliminate the steady-state error or offset.
 Proportional plus derivative control

o Since at steady-state, controlled variable attains new value which is constant.

o Thus, no rate of change of the error occurs and no derivative action.

o Controller output, in this case, equals its nominal value since 𝑑𝜀(𝑡)Τ𝑑𝑡 = 0.

o Hence, derivative control is always combined with P-control or PI-control and never
used by itself.

o When used with P-control, the result is PD-control whose control law is as
𝑑𝜀(𝑡)
𝐶𝑜 = 𝐶𝑂𝑏𝑖𝑎𝑠 + 𝐾𝐶 𝜀 𝑡 + 𝐶𝑂𝑏𝑖𝑎𝑠 + 𝐾𝐶 𝜏𝐷 9.8
𝑑𝑡
𝑑𝜀(𝑡)
𝐶𝑜 = 𝐶𝑂𝑏𝑖𝑎𝑠 + 𝐾𝐶 𝜀 𝑡 + 𝜏𝐷 9.9
𝑑𝑡
 Proportional plus derivative control

o Furthermore, derivative control action requires pure error signal for it to be accurate.

 For process measurements containing high-frequency, random fluctuations, i.e.,

noises.

 The derivative of the error will be wildly changing.

o Derivate control makes controller output worse by amplifying such noises

unless measurement is filtered.

o Derivative control has a tendency to stabilize a controlled process.

o Thus, it is often used to counteract the destabilizing tendencies of integral action.

 Proportional plus derivative control

o It also has a tendency to reduce settling time, i.e., time taken for response to attain
steady state.

o This improves the dynamic response of a controlled process.

o Ideal proportional plus derivative algorithm (Eq. 9.9) is not physically realizable.

 That is, it cannot be implemented in its exact form in analog or digital controllers.

 Actual algorithm implemented contains a derivative filter.

Proportional plus integral plus derivative (PID) control

 PID control is essentially a combination of all the three basic control modes (P, I & D).
 Proportional plus integral plus derivative (PID) control

o It provides the best option that retains all positive attributes of each basic control mode in
one final control algorithm.
o That is, PID control, combines the advantages of all three control algorithms.
o Offset is eliminated by the presence of the integral control action.
o Derivative action on the other hand reduces maximum deviation or overshoot, and
o duration for which oscillations occur in a systems response, i.e., settling time.
o Proportional control provides for immediate action thus PID control law is
𝑡
1 𝑑𝜀(𝑡)
𝐶𝑜 = 𝐶𝑂𝑏𝑖𝑎𝑠 + 𝐾𝐶 𝜀 𝑡 + න 𝜀 𝑡 𝑑𝑡 + 𝜏𝐷 9.10
𝜏𝐼 𝑑𝑡
0
 Proportional plus integral plus derivative control

o Summary of the characteristics of different control modes.

No control

Controlled variable

Offset

PID PI PD P

0 Time
0
Fig.10.1 Response of a controlled variable to a step change in disturbance for different control modes
 On-off control

o With on-off control, the controller output signal has only two possible states and can be
expressed as follows:
𝐶𝑂𝑚𝑎𝑥 𝑖𝑓 𝜀 ≥ 0
𝐶𝑂 = ൜ 9.11
𝐶𝑂𝑚𝑖𝑛 𝑖𝑓 𝜀 ≤ 0

o The controller outputs 𝑪𝑶𝒎𝒂𝒙 and 𝑪𝑶𝒎𝒊𝒏 are the on and off values.

o Therefore, with on-off control, when 𝑪𝑶 = 𝑪𝑶𝒎𝒂𝒙 there is an action from the
controller.

o E.g. turn on the heat supply to a process.

o And when 𝑪𝑶 = 𝑪𝑶𝒎𝒊𝒏 there is an opposite or no action from the controller, e.g.
switch off the heat supply to a process.
 On-off control

o For a digital signal such as used in computer, 𝑪𝑶𝒎𝒂𝒙 = 𝟏𝟎𝟎% and 𝑪𝑶𝒎𝒊𝒏 = 𝟎%.

o On the other hand for analog controller whose output signal is a current,

o output current signal is usually𝑪𝑶𝒎𝒂𝒙 = 𝟐𝟎 𝒎𝑨(on position) and 𝑪𝑶𝒎𝒊𝒏 = 𝟒 𝒎𝑨

(off position).

o Since on-off control has only two-states, it is also known as two-position control.

o Can be considered as a special type of proportional-only control with a large gain.

o That is, when the error is infinitesimal, the output of a controller is zero (off-position)

o and when the error is significant, the output jumps to a large value (on-position).
 On-off control

o On-off controllers are commonly used as thermostats in furnaces and heating systems.

o Furthermore, they are also used in non-critical industrial applications such as some
level control switches.

o But they are not as widely used as PID controllers since they are not as versatile
or as effective.

o On-off control results in continuous cycling of the controlled variable.

o This leads to excessive wear of the final control element.

o Although, this is only significant when the final control element is a control valve and
not so when a solenoid valve is used.