5.

Dynamic Behavior of First-Order

and Second-Order Processes
5.1 Introduction

Dynamic models have been derived for typical processes in lect.2 and standard transfer function form of

these models have been explained in lect.4. The response of different processes to typical changes in their

environment (changes in their inputs; disturbance variables) will be investigated in this lecture.

The transfer function representation makes it easy to compare the effects of different inputs. A second

advantage of the transfer function representation is that the dynamic behavior of a given process can be

generalized easily. Once the response of the process has been analyzed to an input change, the response

of any other process described by the same generic transfer function is then known.

1
 5.2 Standard Process Inputs
In analyzing process dynamic and process control systems, it is important to know how the process
responds to changes in the process inputs. A number of standard types of input changes are widely used
for two reasons:

1- They are representative of the types of changes that occur in plants.

2- They are easy to analyze mathematically.

5.2.1 Step Input

A sudden change in a process variable can be approximated by a step change of magnitude, M:

(1) •Special Case: If M = 1, we have a “unit step change”. We give it

the symbol, S(t).
(2)
•Example of a step change: A reactor feedstock is suddenly
switched from one supply to another, causing sudden changes
The step change occurs at an arbitrary time in feed concentration, flow, etc.
denoted as t = 0.
2
Note that uS(t) is defined as a deviation variable (the change from the normal steady state). Suppose the
heat input to a stirred-tank heating unit suddenly is changed from 8000 to 10,000 kcal/h, by changing the
electrical heater input. Then
Q(t) = 8000 + 2000 S(t) , Q′(t) = 2000 S(t) (*)

5.2.2 Ramp Input

• Industrial processes often experience “drifting disturbances”, that is, relatively slow changes up or down
for some period of time.
• The rate of change is approximately constant.
We can approximate a drifting disturbance by a ramp input:
(3)
(4)
Examples of ramp changes:
Ramp a setpoint to a new value. Feed composition, heat exchanger fouling, catalyst activity, ambient
temperature.

3
5.2.3 Rectangular Pulse
It represents a brief, sudden change in a process variable:

(5)

Equation 5 is depicted in the following Figure as the sum of two steps, one step of magnitude equal to 1
occurring at t = 0 combined with a second step of magnitude equal to −1 occurring at t = tw. Mathematically,
this combination can be expressed as:
uRP(t) = h[s(t)-s(t-tw)]=h[1-s(t-tw)] (6)

Examples:
Reactor feed is shut off for one hour.
The fuel gas supply to a furnace is briefly interrupted.

4
5
Other Inputs:
1- Sinusoidal Input:
Processes are also subject to periodic, or cyclic, disturbances. They can be approximated by a sinusoidal
disturbance:

(7)
where: A = amplitude, ω = angular frequency (ω in radians/time) [where: Period (P)=2π/w]
The Laplace transform of the sine function: U sin ( s ) =
A (8)
s2 +  2

Sinusoidal inputs are particularly important, because they play a central role in frequency response analysis

Examples:
• 24 hour variations (daily changes) in cooling water temperature.
• high-frequency disturbances associated with mixing and pumping operations,
• high-frequency disturbances associated with 60-Hz electrical noise arising from AC electrical equipment
and instrumentation.

6
2- Impulse Input:
Here, U I ( t ) =  ( t ) and U I ( s ) = 1 (9)
• It represents a short, transient disturbance. It is the limit of a rectangular pulse for tw→0 and h = 1/tw
• Exact impulse functions are not encountered in normal plant operations. To obtain an impulse input, it is
necessary to inject a finite amount of energy or material into a process in an infinitesimal length of time,
which is not possible.
Examples:
• Electrical noise spike in a thermo-couple reading.
• Injection of a tracer dye.
So, Y(s)=G(s) [where U(s)=1] (10)

3- Random Inputs:
Many process inputs change in such a complex manner that it is not possible to describe them as
deterministic functions of time. If an input u exhibits apparently random fluctuation, it is convenient to
characterize it in statistical term (to specify its mean value μ and standard deviation σ).

7
 5.3 RESPONSE OF FIRST-ORDER PROCESSES
The general first-order transfer function is given by: G(s)=Y(s)/U(s)= K/(τs + 1) (10)
where: K is the steady-state gain and τ is the time constant.

Some particular forms of input U(s) will be investigated to derive expressions for the response, y(t):

5.3.1 Step Response:

For a step input of magnitude M, U(s) = M/s, and Eqn. 10 becomes Y(s) = KM/[s(τs + 1)] (11)

Taking Laplace Inverse, the time-domain response is:

y(t) = KM(1 − e−t∕τ) (12)

The plot of this response and its normalized values [y/(KM)] are shown below:

8
Soln.
(a)

(b) Q’(s)=0.3*3*107/s=9*106/s
So, T’(t) = (9*106*10-6)[1-e(-t/1.8)]=9[1-e(-t/1.8)]
Note, T’(∞) = 9 oC
So, 0.99(9) = 9[1-e(-t/1.8)]
-t/1.8=ln(1-0.99)=ln(0.01)=-2
t=-2*(-1.8)=3.6 h

(c) T’i(s)= 20/s

So, T’(t) = (20*1)[1-e(-t/1.8)]=20[1-e(-t/1.8)]
135-130=20[1-e(-t/1.8)]=5
So, -t/1.8=ln[1-(5/20)]=ln(0.75)=-0.28768
Then, t=-1.8*(-0.288)=0.52 h

9
5.3.2 Ramp Response:
The response of a first-order system to a ramp input, U(s) = a/s2 will be investigated:
Y(s)=G(s)*U(s)= Ka/[s2(τs + 1)] (12)
Performing a partial fraction expansion yields: Y(s) = α1/(τs + 1)+ α2/s+ α3/s2 (13)
The Heaviside expansion (Chapter 3) gives: Y(s) = Kaτ2/(τs + 1)− Kaτ/s+ Ka/s2 (14)
Using Laplace inverse: y(t) = Kaτ(e−t∕τ − 1) + Kat (15)
The above expression has the interesting property that for large values of time (t ≫ τ):
y(t) ≈ Ka(t − τ) (16)
• Equation 16 implies that after an initial transient period, the ramp
input yields a ramp output with slope equal to Ka, but shifted in time
by the process time constant τ (see the following figure).
• An unbounded ramp input will ultimately cause some process
component to saturate, so the duration of the ramp input ordinarily is
limited.
• A process input frequently will be ramped from one value to another in
a fixed amount of time so as to avoid the sudden change associated
with a step change.
• Ramp inputs of this type are particularly useful during the start-up of a
continuous process or in operating a batch process.
10
5.3.3 Sinusoidal Response:
A sinusoidal input [usin(t) = A sin ωt and U(s)=Aw/(s2+w2)] will be considered as a final example of the response of first-order processes.
So, Y(s)=G(s)*U(s)=KAw/[(τs + 1)(s2+w2)] (17)
Performing partial fraction expansion and Heaviside expansion yields:
Y(s)= [KA/(ω2τ2 + 1)]*[ωτ2/(τs + 1)− sωτ/(s2 + ω2) + ω/(s2 + ω2)] (18)
Taking the inverse Laplace transform gives:

(19)

(20)

(21)

Note: f is not a function of t but of τ and w only.

• The exponential term goes to zero as t → ∞, leaving a pure sinusoidal response as a long-time response/1st order frequency response (y𝓁):
(22)
• Output sine is attenuated (decreased) by angular frequency w: very rapidly/fast oscillation relative to the residence time (w>>1/τ) will lead to a
completely attenuated sinusoidal input so the sinusoidal deviation in the output signal is very small, while very slow oscillation relative to the
residence time (w<<1/τ) will lead to the output is in phase with the input, tracking the sinusoidal input as if the process model were G(s) = K
1
• The long-time response y𝓁(t) is called the frequency response of the first-order system and has two
distinctive features (see the following figure).
1. The output signal is a sine wave that has the same frequency, but its phase is shifted relative to the
input sine wave by the angle ϕ (referred to as the phase shift or the phase angle); the amount of
phase shift depends on the forcing frequency ω.
2. The sine wave has an amplitude  that is a function of the forcing frequency:
(23)

Dividing both sides of Eqn. 23 by the input signal amplitude A yields the amplitude ratio (AR):
(24)

which can, in turn, be divided by the process gain to yield the normalized amplitude ratio (ARN):

(25)

2
 So,
H’(t)= [(KAwτ)/(w2τ2+1)]e(-t/τ) + [KA/ w2τ2+1]sin(wt+Φ)
For t=6min (360s), e(-t/τ) = e(-360/50)=e-7.2<10-3 (very
small term and can be ignored).
The amplitude of the output (level) perturbation is:
KA/ w2τ2+1 =
10(s/m2)(0.1 m3/s)
=0.85m
[(6.28 rad/cycles)(0.002 cycles/s)(50 s)]2+1
Hence, the actual tank level varies from a minimum
of 3.15 m to a maximum of 4.85 m.
For very low frequencies (ω → 0), the largest output
deviations that can result from sinusoidal variations of
amplitude 0.1 m3/s occur. In this case, the deviations
would be ±KA = ±(10 s/m2) (0.1 m3/s) = ±1m. Hence, the
minimum and maximum values of liquid level would be 3
and 5 m, respectively.
For high-frequency variations (0.2 cycles/s), the output
Soln.
amplitude approaches zero. This occurs because the
Note: K=10, τ=50, w=0.002*(2π)=0.01256 rad/s rapid variations of flow rate are averaged in the tank
Φ=-tan-1(wτ)=-tan-1(0.01256*50)=-0.56 or -32o when the residence time is sufficiently large, giving a
A=0.1 m3/s, Q’(t)=Asin(wt)=0.1sin(0.01256t) relatively constant level.
3
 5.4 Response of Integrating Processes
The integrating process is described in the liquid storage tank (case: const. q) {see Lec.4}. The outflow rate
q can be set at any time by adjusting the speed of the pump (q will be an additional input/disturbance
variable):
Adh/dt = qi - q (26) , at steady state: qഥi = qത (27)
Adh’/dt = q’i - q’ (28) , Taking Laplace: AsH’(s) = Q’i(s) – Q’(s) (29)
1
so, H’(s)= 𝐴𝑠[Q’i(s) – Q’(s)] (30) ,
Both transfer functions, H′(s)/Q′i (s) = 1/As and H′(s)/Q′(s) = −1/(As), represent integrating models,
characterized by the term (1/s). The integral of Eqn. 26 is:
1 𝑡
h(t) − hത = ‫׬‬0 [qi(t∗) − q(t∗)]dt∗ (31) {Integrating process}
𝐴
Integrating processes do not have a steady-state gain in the usual sense.
Any positive step change in qi (increase in qi above q) will cause the tank level to increase linearly with time
in proportion to the difference, qi(t) − q(t)
Any positive step change in q will cause the tank level to decrease linearly.
Thus, no new steady state will be attained, unless the tank overflows or empties. In contrast, a tank with an
exit line valve, rather than a pump, will reach a steady state when the outflow rate becomes equal to the
inflow rate. This process is described by a first-order transfer function rather than an integrator (see Lect.4).
4
• Integrating systems of step input functions represent an example of non-self-regulating processes.
Arbitrary pulse inputs do lead to a new steady state. For example, the response of an integrating process
of the rectangular pulse input will be as below:
Y(s)/U(s) = K/s (32) , Y(s) = Kh(1-𝑒 −𝑡𝑤 𝑠 )/s2 = Kh[(1/s2)-(1/s2)𝑒 −𝑡𝑤 𝑠 ] (33a)
So, y(t) = Kh[ts(t)-(t-tw)s(t-tw)] (33b)
So, y(t) = Kht for t<tw and y(t) = Khtw for t≥tw (34)
Equation 34 shows that the change in y at any time is proportional to the area under the input pulse (the
integral), an intuitive result.

5
6
 5.5 RESPONSE OF SECOND-ORDER PROCESSES
A second-order transfer function can arise physically whenever two first-order processes are connected in
series (such as the 2 stirred-tank blending processes, see Lect. 4).
𝑌(𝑠) K1K2 K
G(s) = 𝑈(𝑠) = (τ 𝑠+1)(τ = (35)
1 2
𝑠+1) (τ 𝑠+1)(τ 𝑠+1)
1 2

Also, a second-order process transfer function will arise upon

transforming a second-order differential equation process model
such as the two coupled first-order differential equations for the CSTR (as below):
VdcA/dt = q(cAi − cA) − VkcA (36) , VρCdT/dt = wC(Ti − T) + (−ΔHR)VkcA + UA(Tc − T)
(37) where: k = k0e−E∕RT (38)
So, dc′A/dt = a11c′A + a12T′ (39) , dT′/dt = a21c′A + a22T′ + b2T′c
(40a)
Taking Laplace Transformation: (41)
(42)
Substituting for T’(s), eqn. 41 becomes: (43)
Eqn. 44 is a second-order transfer function. The
a and b coefficients can be evaluated for (44)
specified nominal conditions.
(40b) 7
Second-order transfer functions will be considered in the standard form:
𝐾
G(s) = (45)
τ2s2 + 2ζτs + 1
Where:
K is the steady state gain
τ determines the system response speed (or time)
ζ (zeta) is damping coefficient and dimensionless. It provides a measure of the amount of damping in the
system (the degree of oscillation in a process response after an input change).
Note: sometime 2nd order system is simplified to 1st order system (simplest model) such as an electrically
heated stirred tank Model in Lect.2:

(46) y = T −T u=Q-Q Ti fixed (49)

(47)
mm e Ce d 2 y  m e Ce m e Ce m  dy 1
2
+ + +  +y=  u (50)
wh e A e dt  h eAe wC w  dt wC
note when Ce → 0, obtain 1st order equation (simpler model)
(48)
8
 (eqn. 45)

(by equate eqns. 45 and 51)

(52) (51) (from eqn. 35)

(53)

(54) SOLUTION
From Eqns. 52 and 53:
• For process control problems, Eqns. 55 and 56 provide a check on
the underdamped form is these results:
(55) frequently encountered in
investigating the properties of
(56)
processes under feedback
control.
9
5.5.1 Step Response of 2nd order system:
For the step input (U(s) = M/s) to a 2nd order process described by Eqn. 45:
𝐾𝑀
Y(s) = (57)
s(τ2s2 + 2ζτs + 1)
After inverting to the time domain, the responses can be categorized into three classes:
Overdamped (𝛇 > 1):
If the denominator of Eqn. 57 can be factored using Eqns. 55 and 56, then the response can be written:
τ1exp(−t/τ1)− τ2exp(−t/τ2)
y(t) = KM[1 − ] (58)
τ1−τ2

(59)

(60)

10
When damped oscillation is desirable, values of ζ
in the range 0.4 to 0.8 may be chosen. In this
range, the controlled variable y reaches the new
operating point faster than with ζ = 1.0 or 1.5, but
the response is much less oscillatory (settles
faster) than with ζ = 0.2.
11
The following terms are used to describe the dynamics
of underdamped processes:
1. Rise Time (tr) is the time the process output takes
to first reach the new steady-state value.
2. Time to First Peak (tp) is the time required for the
output to reach its first maximum value.
3. Settling Time (ts) is the time required for the process
output to reach and remain inside a band whose
width is equal to ±5%of the total change in y (±1%
is also used for some applications).
4. Overshoot (OS) = a/b (% overshoot is 100 a/b).
5. Decay Ratio (DR) = c/a (where c is the height of the
second peak).
6. Period of Oscillation (P): is the time between two
successive peaks or two successive valleys of the
response.
Note: The above definitions apply more generally
to the step response of any underdamped process. If
the process does not exhibit overshoot, the rise time
definition is modified to be the time required to go
from 10% to 90% of the steady-state response
12
 For the particular case of an underdamped second-order process, analytical expressions for some of these
characteristics can be developed using eqn. 60:

(61)

(62)

= (OS)2 (63)

(64)
Note:
• OS and DR are functions of ζ only. For a second-order system, the
decay ratio is constant for each successive pair of peaks. Figure 5.10
illustrates the dependence of overshoot and decay ratio on damping
coefficient.
• For an underdamped second-order transfer function, either graphical
approach using [figs. 5.7 and 5.10 and eqn. 64] or analytical approach
using [eqns. (61 or 62) and eqn. 64] can be used to obtain estimates of
ζ and τ based on step response characteristics.
13
Soln.

(a)
The gain is obtained by dividing the steady-state change in
temperature by the feed flow rate (disturbance) change:
K = (102 − 100)/(0.5 − 0.4) = 20 ∘C/(kg/s)

(b)
The oscillatory characteristics of the response [graphically using
figures (5.7 and 5.10) or analytically using either eqn. 61 or eqn.
63] can be used to find the dynamic elements (ζ and τ) in the
transfer function relating temperature to feed flow rate.
Using eqn. 45, G(s) = K/(τ2s2 + 2ζτs + 1)
Analytical Approach to Estimate ζ and τ:
OS = (102.5 − 102)/(102 − 100) = 0.5/2 = 0.25 (i.e., 25%)
[ln(OS)]2
Re-Arranging eqn. 61: ζ = = 0.4037 ≈ 0.4
π2 + [ln(OS)]2
P= 3060 − 1000 = 2060 s
1 − ζ2
Re-Arranging eqn. 64: τ = P = 300 s
2π
G(s)=T′(s)/W′(s)=20/[3002s2+(2*0.4*300)s+1]=20/[90000s2+240s+1]

14
(c)
The rise time tr can be calculated from Eqn. 60. When t = tr, y(t) is equal to its final steady-state value (KM). In
other words, the bracketed quantity is identically zero at t = tr:

(65)

The general solution has multiple values of t that satisfy y(t) = KM:

(66)

The rise time corresponds to the first time (n = 1) that y(t) = KM = y(∞). Solving for the rise time gives:

(67)

Where: the result of the inverse cosine computation must be in radians.

Because τ = 300 s and ζ = 0.40, using eqn. 67: tr = 649 s

15
5.5.2 Sinusoidal Response of 2nd order system:
When a linear second-order system is forced by a sinusoidal input A sin ωt, the output for large values of time (after exponential
terms have become negligible) is also a sinusoidal signal given by:

(68)

(69)
KA
• The output amplitude  ̂ is obtained directly from Eqn. 68:  = (70)
[1 − (ωτ)2]2 + (2ζωτ)2
• The ratio of output to input amplitude is the amplitude ratio AR (= Â/A). When normalized by the process gain, it is called the
 1
normalized amplitude ratio ARN: ARN = =
KA [1 − (ωτ)2]2 + (2ζωτ)2
(71)

• ARN represents the effect of the dynamic model parameters (ζ, τ) on the sinusoidal response; that is, AR N is independent of
steady-state gain K and the amplitude of the forcing function, A. The maximum value of AR N (if it exists) can be found by
differentiating Eqn. 71 with respect to ω and setting the derivative to zero. Solving for ωmax gives:

1 − 2ζ2
ωmax = (for 0 < ζ < 0.707) (72) [For ζ ≥ 0.707, there is no maximum, as Fig. 5.11 illustrates]
τ
Substituting Eqn. 72 into Eqn. 71 yields an expression for the maximum value of ARN:
(73)
• Eqn. 73 indicates that a process with little damping can exhibit very large output oscillations
if it is perturbed by periodic signals with ω near ωmax.
16
17
 τ and ζ are calculated from Eqns. 52 and 53:
τ = (3)(10)= 5.48 s
ζ = 132(5.477) = 1.19
The frequency of the perturbing sinusoidal signal (reactor
temperature) is calculated from the observed period of 30 s:
ω = 2π/P= 6.28/30= 0.2093 rad/s
The amplitude  of the output perturbation also is obtained
from observed results as
 = (183 − 180)/2 = 1.5 ∘C
Equation 70 now can be rearranged to calculate the
amplitude of the actual reactor temperature
A =(Â/K) [1 − (ωτ)2]2 + (2ζωτ)2 = 4.12 ∘C
Soln. Thus, the actual reactor temperature is varying between
First, note that the sensor/transmitter and the transmission 181.5 − 4.12 = 177.38 ∘C and 181.5 +4.12 = 185.62 ∘C,
line act as two first-order processes in series (Eq. 35) with nearly three times the variation indicated by the measured
an overall gain K = 1, τ1 = 3 s and τ2 = 10 s value.
So, the approximate transfer Function: Because the second-order process in this example is
G(s)= T′meas(s)/T′reactor(s) = 1/[(3s + 1)(10s + 1)] overdamped (ζ = 1.19), we expect that sinusoidal
The output for large values of time is also a sinusoidal signal perturbations in the reactor temperature always will be
given by eqn. 68 and the output amplitude  is obtained attenuated (reduced in amplitude) in the measurement
using eqn. 70 as mentioned before. system regardless of the frequency of the perturbation.

18