Heat Flow Experiment

System identification & Frequency domain design

1 Objective

Perform system identification on the heatflow experiment and design a controller in the frequency
domain to control the air temperature at selected points in a duct using a heating element and a
blower.

This heat-flow experiment (HFE) system consist of a duct equipped with a heater and a blower at
one end and three temperature sensors located along the duct. The power delivered to the heater is
controlled using an analog signal. The fan speed can also be controlled using an analog signal. Fast
settling platinum temperature transducers are used to measure the temperature. Fan speed is
measured using a tachometer and can be used to design speed controllers.

Figure 1 Heatflow apparatus

2 Open loop model

A complete thermodynamic model of the system is difficult to derive. In principle however the following equation applies:

Where
 Tn Temperature at Sensor n

Vb Voltage applied to blower

Vh Voltage applied to heater

Ta Ambient temperature

Xn Distance of sensor n from heater

This experiment is aimed at introducing you to the system in order to gain insight into the salient issues. You will develop a
very simple first order model and an associated controller. In subsequent experiments you will investigate system
identification techniques and control of systems with time delay.

3 System identification experiment

In the first experiment, we developed an open loop model by examining the step response of the system. In this
experiment, we shall investigate the use of system identification techniques to develop a higher order model and use it to
design feedback controller. Note that we have 5Hz low pass filters to remove some of the noise in the sensors.

Quanser Consulting 2*pi*5

20
MQ3 ADC s+2*pi*5
T emp/Volts1 Transfer Fcn T0

ADC T1

Heat
Select Quanser Consulting 2*pi*5
20
MQ3 ADC s+2*pi*5
Subsystem Temp/Volts2 Transfer Fcn1 T1

0 Quanser Consulting ADC T2

MQ3 DAC
VQ1
Multiport
Switch Analog Output0 2*pi*5
Quanser Consulting
20
MQ3 ADC s+2*pi*5
[Tus' u]
Temp/Volts3 Transfer Fcn2 T2
From Flow
Workspace ADC T3

1200

VF1
Multiport
VF -K- Switch1

VF Speed

In1 Out1

Saturation1 Fan RPM1

Blower
2500

Min

Figure 2 Simulink model to apply random inputs and collect data.

3.1 Data Collection

Follow these steps to collect the data :

Start Matlab
Change the directory where you have the FHE models
In Matlab type random_input - This opens the simulink model shown in Figure 2
In random_input.mdl Click WinCon / Set WinCon Options - This sets the compile for the OS you have.
In random_input.mdl Click WinCon / Set WinCon Options - This sets the compile for the OS you have.
In Matlab run setup_random - This initializes parameters so you can compile the diagram- It also sets up the
random sequence to be applied to the heater
 In the diagram click WinCon/ Build - this will build the realtime code and download to WinCon
In WinCon click on Plot / New / Scope and select T0, T1 T2
In the WinCon plot, select Update/ Buffer and enter 60
Click run - this will run the controller for 240 seconds. (4 minutes!)
When the run is complete , you will see the last 60 seconds of data in the WinCon plot. Click File/ save and save
as a mat file in the file named : data_rand1.mat
Now that the data is saved you can load into Matlab

> load dat_rand1

Plot the data

> plot( plot_time , [random_input_T1,random_input_Heat]);

This shows the plot in Figure 3

The blue trace is the temperature at sensor S1 and the green trace is the random input applied to the heater.

60

50

40 Temp S1
VQ
30

20

10

0
0 20 40 60
Figure 3 Response of T1 to random input VQ. Note that the data
is collected after running the experiment for 4 minutes in order to
remove trends.
3.2 Model estimation

Now that the data is saved, you can run the script id_freq.m - This script loads the data that was collected and uses pem (
Prediction error estimate - See Matlab System Identification toolbox) to create a best fit linear model of the system.

Measured Output and Simulated Model Output

3
Measured Output
m1 Fit: 73.97%
2

1
Temperature - S1

-1

-2

-3
30 35 40 45 50 55 60 65

Figure 4 Simulated versus estimated response from random

input data

Figure 4 shows the actual data versus data from the estimated model parameters.

Figure 5 shows the predicted step response of the system

Open Loop Model - Step Response

6
Amplitude

0 5 10 15
Time (sec)

The estimated open loop transfer function ( for S1) is obtained as:
4 Control System Design

We design controller in the frequency domain. Clearly we require an integrator in the loop in order to obtain zero steady
state error.

The controller we design is of the form:

which consists of a an integrator followed by a lead centred at wm and a gain K.

The design file d_freq uses the estimated parameters and prompts the user for gain and phase lead values to design an
appropriate compensator. (The values in parentheses are the recommended values to enter)

Using d_freq we obtain the following compensator:

5 Setting up

Wire up the system as described below. Use single RCA cables for each connection.

Inputs to system Attach to Analog outputs Range

Vh - Voltage to heater MultiQ D/A # 0 - Control heat delivered 0 - 5 VDC

Vb - Voltage to blower MultiQ D/A # 1 - Control fan speed ( open loop) 0 - 5 VDC

Measurements Attach to Analog inputs

S1 MultiQ A/D #0 0 - 5 VDC

S2 MultiQ A/D #1 0 - 5 VDC

S3 MultiQ A/D #2 0 - 5 VDC

Vt MultiQ A/D #3 0 - 5 VDC

6 Implementation and results

The controller is implemented using WinCon and Simulink. The Simulink controller is in the file q_hfe_freq.mdl The
controller measures the voltages from the three sensors and calibrates them to degrees. The feedback controller
implements the equation.

Note that we have also added a feedforward component Vff which pre-computes the steady state voltage required for the
desired heat. The value of Vff is obtained as follows:

Consider the estimated open loop transfer function:

The steady state gain is G_ss = 1.7/ 0.2 = 8.5 deg/volt. That means that in the steady state, in order to maintain 8.5 deg
change from the nominal value we need 1 volt applied to the heater. We can use this gain to predict the steady state value
required to maintain the desired temperature.

The feedforward voltage to the heater then is :

Depending on the value of the selector (Constant named Sensor in the diagram), we will control one of the three points
along the duct. Note that we are also running a realtime simulation of the system modeled as obtained from the PEM
estimator.

By setting Sensor = 1, we control T1 and make it track a square wave of 5 deg p-p centered about 40 degrees. The
response is shown in Figure 7.

Examining the response we note the following:

1 The output tracks the desired temperature command with zero steady state error
2 The transient response matches the simulation closely

40

Constant -K-
TCmd
FeedForward
Signal
Generator 1 -n
s d(s)
Integrator Transfer Fcn1
1
Saturation2
COMPENSATOR
Sensor

T1

T1
Heater T2

T3 T2
Quanser Consulting
VFS
MQ3 DAC
Heater and Sensors Multiport
Fan Speed Saturation1
Switch
T3
Analog Output1

COMPENSATOR
1 -n
s d(s)
Integrator1 Transfer Fcn2

n_ol(s) Saturation3

d_ol(s)
Transfer Fcn TSim

SIMULATION OF SYSTEM
Figure 6 Feedback controller with frequency domain designed compensator. Note a simulation of the system is running in
tandem.
 45
Command Deg.
44 S1 - Deg
Simulation Deg.
43

42

41

40

39

38

37

36
0 10 20 30 40 50 60

Figure 7 Response at S1 using compensator designed with

estimated parameters- Green trace is actual system response
while red trace is the simulated response using estimated model.

We can easily examine the system response if we try to control the temperature at S3. We do this by setting Sensor = 3 in
the diagram and running it again. In this case the response is shown in Figure 8. We note that the performance
deteriorates, primarily due to the transport delay in the system.

45
Command
44 S3
Simulation
43

42
TEMP - DEG C

41

40

39

38

37

36
0 10 20 30 40 50 60
TIME

Figure 8 Response at S3 when using controller designed for S1

3 Conclusions

The system identification technique results in a good model that can be used to design a frequency domain compensator.
The actual system response closely matches the predicted response from the estimated model. Closing the loop at S3
using the controller designed for S1, although inferior, still has acceptable performance.
4 System Parameters

Parameter Symbol Value Units

Heater maximum power (5 V Input) P 400 Watt

Blower nominal airflow (4 V Input) B 32.1 CFM

Blower nominal airflow (4 V Input) BSI 0.91 m3 / min

Maximum blower speed ( 5 V Input) Fmax 2580 RPM

Minimum blower speed ( 0 V Input) Fmin 1580 RPM

Tachometer calibration Ct 1063 RPM / V

Temperature Sensor calibration Cs 20 Deg C / V

Duct cross section area A 0.0064 m2

Max wind speed Ws 142 m/ min

Temp sensor settling time at max wind speed Ts 4 seconds

Dimensions Power Requirements

50 x 15 x 10 cm VAC 100 - 240

0.5 Kg Amps Max 5

Data Acquisition Requirements

Analog inputs 4 0 - 5 V DC

Analog outputs 2 0 - 5 V DC

5 Calibration

Occasionally you may need to calibrate the sensor measurements. You will need a measurement of ambient temperature
and a DC Voltmeter. (Or you can use WinCon)

Power up the system

Open the service panel as shown
Measure S1 and adjust potentiometer P1 such that the voltage is proportional to room temperature
Measure S2 and adjust potentiometer P2 such that the voltage is proportional to room temperature
Measure S3 and adjust potentiometer P3 such that the voltage is proportional to room temperature

The table below shows the voltages you should obtain at various temperatures:

Temp 19 20 21 22 23 24 25 26 27 28 29 30

Vs 0.95 1.00 1.05 1.10 1.15 1.20 1.25 1.30 1.35 1.40 1.45 1.50