EXPERIMENT-1

Simulate and compare different identification methods for SISO

systems

DEPARTMENT OF CHEMICAL ENGINEERING & ELECTRICAL

ENGINEERING

Submitted By
XXXXXXX
(Roll. No.: XXXXXXX)

1|Page
Objective:
Identification of Model using Graphical Techniques and System Identification tool
box for Single Input-Single Output Systems.
Theory:
System Identification:
Identification is used for developing a mathematical relationship (model) between
the causes (inputs) and the effects (outputs) of a system (process) based on
observed or measured data. Stated otherwise, identification establishes a
mathematical map between the input and output spaces as determined by the data
as depicts in figure shown below:

Process Description:
Calculation of gain, time constant and time delay of the standard first order system
is taken place with the help of three graphical techniques by providing or
manipulating the heat input to the system for controlling temperature of the tank.
Procedure:

Simple Process Model is given with first order system with gain (K), dead
time/delay time (θ) and time constant (τ):

2|Page
Generally, there are three accepted graphical techniques for determining first order
system parameters. In this Process we are obtaining parameters by following
graphical methods:

Graphical Techniques:

1. 63.2% Response Method

2. S&K Method
3. Point of Inflection Method

63.2% Response Method-

Gain(K); Gain of the system is defined as the ratio of change in output to the
change in input.

Time constant (τ): It is defined as the time taken by the response of the system to
reach 63.2% of the final value.

Time delay (θ): Time taken for the system to respond to a change in input for the
first time.

S&K Method:

From SK method (in which time corresponding to 35% and 85% of response is
considered) we have standard equations for time delay (θ) and time constant (τ) of
the system are given as follows:

θ = (1.3*t1 – 0.29*t2)

τ = 0.67*(t2 – t1)

Where, t1 and t2 are the times corresponding to the 35% and 85% of the response
value (excluding input delay) respectively.

3|Page
Point of Inflection Method:

Time Constant (τ) & Time Delay (θ):

A tangent is drawn at the point of inflection of the step response; the intersection of
the tangent line and the corresponding time at time axis is the time delay.

If the tangent is extended to intersect the steady state response

line, the point of intersection corresponds to time,

t=θ+τ

Therefore, τ=t–θ

MATLAB code:

close all
clc
load('CSK_TC_Data_399_Samples_ISA.mat') % load the 400 data samples
H1 = h1s(1:120); % H1 is taking as heat input having 120 data samples
T1 = t1s(1:120); % T1 is the Temperature output having 120 data samples
T = time(1:120); % T is taking as the time in seconds having 120 data samples
figure
subplot(2,1,1)
plot(T,H1, 'b','Linewidth',1)
subplot(2,1,2)
plot(T, T1,'r','Linewidth',2)
grid on
grid minor
xlabel('Time', 'FontWeight','bold')

Results and Analysis:

Input-Output Characteristics wrt Time:

Here, we are taking 120 samples value for heat input which is the step input with
respect to time as shown in fig-(i), and by manipulating this heat input, temperature

4|Page
to be controlled and observing the response characteristic with respect to time as
depicts in fig-(ii).

Fig-(i) Heat Input v/s Time

Fig-(ii)Temperature Output v/s Time

5|Page
 Fig-(iii) Temperature(°C) output v/s Time(sec)

Observations & Calculations:

 For 63.2% Response:

63.2% of Change in Output = 0.632*(Final value of o/p – Initial value of o/p)

= 0.632*(57.4219 – 26.17)

= 0.632*31.25 = 19.75°C

As the response is not zero at t = 0, there is a base value (26.17 from graph) which
is to be added in the existing value of response hence, the response becomes,

= 26.17 + 19.75 = 45.92°C

Now, the corresponding time (from fig-(ii) on x-axis) for 45.92°C value of
response would be, t = 68.3 sec

From the Figure (i) and (ii),

6|Page
Time delay (θ) = Output delay – Input delay = 19.09 – 3.01 = 16.08 sec

Therefore, Time constant (τ) = t – θ = 68.3 – 16.08 = 52.22 sec

Gain (K) = Change in output / Change in Input

K = (57.4219 – 26.17) / (80 - 0)

Gain (K) = 0.3906

Hence, the transfer function model will be,

T1(S)/H1(S) = (0.3906/1+52.22 s)*e-16.08s

 For SK method:

Response 1= Base value + 35% of change in response

= 26.17 + 0.35*31.25 = 37.107

Response 2 = Base value + 85% of change in response

= 26.17 + 0.85*31.25 = 52.73

As there is 3 second of delay in the input characteristics as seen from fig-(i),

Hence, t1 = 44.22 – 3 = 41.22 sec

And, t2 = 91.45 – 3 = 89.45 sec

Now, Time constant, τ = 0.67*(t2 – t1) = 0.67*(89.45 – 41.22) = 32.31 sec.

Time Delay, θ = (1.3*t1 – 0.29*t2) = (1.3*41.22 – 0.29*89.45) = 27.64 sec.

Hence, the transfer function model from method-2 will be,

T1(S)/H1(S) = (0.3906 /1 + 32.31 s) *e-27.64 s

7|Page
  For Point of Inflection Method-

From fig-(iii), t = 84.42 sec and θ = 16.08 sec

Hence, Time constant, τ = t – θ = 84.42 – 16.08 = 68.34 sec

Gain (K) = Change in output / Change in Input

= (57.4219 – 26.17) / (80 - 0) = 0.3906

Hence, the transfer function model from method-3 will be,

T1(S)/H1(S) = (0.3906 /1 + 68.34 s)*e-16.08s

8|Page
Process Description for SYSID tool box:
Performing the estimation and validation of data and selecting different sample
ranges by considering system delay as 11.1 seconds and also without delay with
the help of System Identification tool box.
Procedure:
Step I: Create data object from experimental data by using the syntax given below:
 data1=iddata (y,u,ts)
where, y=output data vector,
u= Input data vector,
ts = sampling time.
We have variables as: y = t1s ; u = h1s ; ts = 1.005 sec
Step II: Open system identification toolbox from apps in MATLAB and give the
command: systemIdentification
Step III:
 Export data object in system identification toolbox.
 Import data by selecting ‘Time domain data’ and then select workspace
variable
 Input = h1s, output = t1s, start time = 0, and sample time = 1.005 sec. and
then do Import.
Step IV:
 Pre-process the data by ‘Select Range’.
 In which we take samples for estimation as well as for validation.
Step V:
 Estimate the model.
 In which, we drag our estimation samples graph (having suffix ‘e’ in the
data name shows in data views) to working data for Estimation and
validation samples graph (having suffix ‘v’ in the data name, shows in data
views) to the Validation data.

9|Page
  Select ‘Transfer function model’ and enter number of poles and zeros with
the time delay.
 And by clicking on the graph obtained, we have the transfer function and
values of FPE, MSE, and best estimation fit.
Step VI:
Check the model output best fit by dragging our reference data into estimation as
well as into validation.
Step VII:
Export the model which is best among all estimated models by checking fitness
and errors.
Step VIII: Design controller to the Exported model.

10 | P a g e