How to invert an output error model?

Hi matlab community

I am looking for a way to invert an output error model so that the original output data becomes the input data to predict the original input data.

Best regards

Benjamin

[Sam Chak]

Hi @Benjamin Pommer

So, you want to invert a "static" model,

to become so that you can determine x from the data y?

[Benjamin Pommer]

yes, kind of. It would be a idpoly function which contains a parameter to describe the noise in the system. But, I guess its not working since the degree of the denominator (=2) is greater than the one of the numerator (=1) with a dead time of 0.

The function finverse didnt work. Do you have another idea maybe?

Answer 1

[Rishav]

Hi Benjamin,

I understand that you are trying to invert an output error model with an idpoly function, where you can use the output data to predict the original input data.

• Assuming you have the original output error (OE) model:

original_model = idpoly(...); % Define your original output error model

• Generate simulated data: You need to get simulated output data from the original OE model.

% Simulate output data from the original model
simulated_output_data = sim(original_model, input_data); % Replace input_data with your input data
% Create an iddata object with only output data
Ts = 0.1; % Associated sample time
original_data = iddata(simulated_output_data, [], Ts);

• Invert the model: You can use the original data and the original OE model to estimate an inverted model using the 'oe' function.

inverted_model = oe(original_data, original_model);

• Now since, inverted_model is your inverted model, and you can use it to predict the original input data from the original output data as follows:

predicted_input = predict(inverted_model, original_data.OutputData);

Moreover, you can also refer to the below mentioned documentation links to know more about 'idpoly', 'iddata' and 'oe' functions:

https://www.mathworks.com/help/ident/ref/idpoly.html

https://www.mathworks.com/help/ident/ref/iddata.html

https://www.mathworks.com/help/ident/ref/oe.html

Thank you,

Rishav Saha