Please help: Real time control results difference with simulation

Thread Starter

YukiWong

Joined Dec 5, 2011
27
Hi all,

I had done design and implement a PI controller by using MatLab which give me an overshoot of 25.8% and rise time of 0.09s.
Unfortunately, i had validate this controller in real-time control through Labview and gave me difference results as from simulation where the overshoot is 2.9% and rise time is 0.06s.
At 1st i told there is a scale between real time control and simulation, so i decide to work out other controller. Unfortunately, the results dint help me but make me more confused.
Below is the result of other PI controller with Ziegler Nichols method.
With Z-N open loop method, from simulation i got 71.6% but in real time i got 12.4%.
With Z-N close loop method, from simulation i got 56% but in real time i got 10.4%.
Could anyone explain to me the problem i face please?

Thankyou very much.
 

Thread Starter

YukiWong

Joined Dec 5, 2011
27
It would be helpful if you post some clearer details of the 'plant' transfer function and the PI control parameters.
It is an first order plus dead time system with
gain=58.46
time constant=0.55
time delay=0.06
The plant transfer function is 58.46/(0.55*s+1)*e^-0.06s

for the PI control parameter,
Gc=Kc(1+(1/Ti*s))
For the 1st method with 25.8% overshoot, Kc=0.0784 while Ti=0.24;
For ZN open loop, Kc=0.1441, Ti=0.1998
For ZN close loop, Kc=0.1170, Ti=0.1908

Thankyou for your reply :)
 

Thread Starter

YukiWong

Joined Dec 5, 2011
27
How did you approximate dead-time in your simulations - with a Pade function?
I dint do it in pade function, i just do it with the code below in Matlab.
s = tf('s');
Kss = 58.46;
tao = 0.55;
G = tf(Kss,[tao 1],'InputDelay',0.06);
 

Thread Starter

YukiWong

Joined Dec 5, 2011
27
Consequently, this is the step response to the controller by using simulation.



and

This is the result from Real time control

 

t_n_k

Joined Mar 6, 2009
5,455
You mention the real-time validation was done using Labview. What did that involve - what are the details of the procedure you used?
 

Thread Starter

YukiWong

Joined Dec 5, 2011
27
You mention the real-time validation was done using Labview. What did that involve - what are the details of the procedure you used?
I just do it in simple way through Labview 2011 control and design simulation toolbox, picture below show the regarding Labview Vi block diagram.
I had used convert to dynamic data and convert from dynamic data to convert the controller output into dynamic data before and after DAQ Assistant.



Is the picture helps to understand how my real-time validation been done?

Thanks for your reply,thankyou very much :)
 

t_n_k

Joined Mar 6, 2009
5,455
While I understand what you are doing I'm a little puzzled ....

You are simulating a DC drive which presumably has the same plant transfer function you gave earlier. Is the DAQ assistant providing the interface for driving a physical DC motor with shaft tachometer feedback? Are you sure the transfer function corresponds to actual system?

Also you seem to be validating a PID control method whereas your original investigation was based on a simpler PI control system.
 

Thread Starter

YukiWong

Joined Dec 5, 2011
27
While I understand what you are doing I'm a little puzzled ....

You are simulating a DC drive which presumably has the same plant transfer function you gave earlier. Is the DAQ assistant providing the interface for driving a physical DC motor with shaft tachometer feedback? Are you sure the transfer function corresponds to actual system?

Also you seem to be validating a PID control method whereas your original investigation was based on a simpler PI control system.
Yea, thats is what i gave earlier.
Yes, it is a DC motor with tachometer feedback.
Yes, they are interface through the I/O port of the Labview NI card.
The transfer function is not really correspond to the actual system, because the gain for different value of voltage input are difference. I am just taking the average gain,time constant and time delay while doing model development to improve the accuracy. It is not 100% correspond to the real plant.

It is a parallel PID controller, initially the Td is equal to zero will equal to the PI controller.

thanks for your reply,thank you :)
 

t_n_k

Joined Mar 6, 2009
5,455
Well if the physical system transfer function bears no relationship to the original model, you can hardly expect the physical system response to bear any resemblance to the one you have calculated in your modelling.

You need to think through your concepts more carefully - preferably before posting your questions on the forum.
 

Thread Starter

YukiWong

Joined Dec 5, 2011
27
Well if the physical system transfer function bears no relationship to the original model, you can hardly expect the physical system response to bear any resemblance to the one you have calculated in your modelling.

You need to think through your concepts more carefully - preferably before posting your questions on the forum.
Okay, i got it. Thanks for your advice and reply. Thankyou very much :)
 
Top