Prev: how to do a running max efficiently in Matlab?
Next: finding the four corners pixel location values of a small reectangular image in a large image
From: Daniel on 18 Jan 2010 22:22
I have a state space model of a thermal system for which I am trying to design a temperature estimator. There are a few parameters within my state matrix that are difficult to calculate offline very percisely, and for which I currently have estimated values as place holders.

I also have test data with which to correlate my model.

Is there a good way in Matlab to tune un-known state parameters to match the test data? I know approximatley what they should be and I would like a way to best match the current model to the test data keeping the uknown parameters within a given (physically reasonable) range.

Thanks
From: Rajiv Singh on 2 Feb 2010 17:26
See linear/nonlinear grey box modeling feature in System Identification
Toolbox.
Basically, you can write a MATLAB file containing your equations (your
state-space equations; the file should return the state matrices as a
function of your parameters which will be the inputs to the file) and use it
to create what is called a "grey box model". This model then lets you define
parameters and estimate their values using data. See:

http://www.mathworks.com/access/helpdesk/help/toolbox/ident/ug/bqs6lf8.html

http://www.mathworks.com/products/sysid/demos.html

The demo iddemo7.m ("
Building Structured and User-Defined Models Using System Identification
Toolbox(TM)

") should be particularly useful; on web at:
http://www.mathworks.com/products/sysid/demos.html?file=/products/demos/shipping/ident/iddemo7.html

Rajiv

"Daniel " <djberry(a)oakland.edu> wrote in message
news:hj38gt$2pk$1(a)fred.mathworks.com...
>I have a state space model of a thermal system for which I am trying to
>design a temperature estimator. There are a few parameters within my state
>matrix that are difficult to calculate offline very percisely, and for
>which I currently have estimated values as place holders.
>
> I also have test data with which to correlate my model.
>
> Is there a good way in Matlab to tune un-known state parameters to match
> the test data? I know approximatley what they should be and I would like
> a way to best match the current model to the test data keeping the uknown
> parameters within a given (physically reasonable) range.
>
> Thanks


From: Bora Eryilmaz on 5 Feb 2010 16:24
You can use the System Identification Toolbox's grey-box (idgrey)
identification method to estimate the elements of a linear state-space
model.



"Daniel " <djberry(a)oakland.edu> wrote in message
news:hj38gt$2pk$1(a)fred.mathworks.com...
>I have a state space model of a thermal system for which I am trying to
>design a temperature estimator. There are a few parameters within my state
>matrix that are difficult to calculate offline very percisely, and for
>which I currently have estimated values as place holders.
>
> I also have test data with which to correlate my model.
>
> Is there a good way in Matlab to tune un-known state parameters to match
> the test data? I know approximatley what they should be and I would like
> a way to best match the current model to the test data keeping the uknown
> parameters within a given (physically reasonable) range.
>
> Thanks
>


 | 
Pages: 1
Prev: how to do a running max efficiently in Matlab?
Next: finding the four corners pixel location values of a small reectangular image in a large image