How to find the model of a black box with a known input and output signature of signal.

Please help me identify the model of a black box with a known input x(t) and output y(t), using the provided DATA.mat.

I have attempted to map the output to the given input using MATLAB System Identification Toolbox and neural networks, as described in the application section, but have not been successful.

Kindly assist in mapping this input to the output. You are free to use any methods.

Attacked input and output data.

[Star Strider]

Wheen I plot the second column of ‘data’ as a function of the first, the result is simply a dense series of horizonotal lines. I can’t make any sense of it.

[CM Sahu]

It's a sine wave with a frequency close to 17.5 kHz. If you zoom in, you'll be able to see the waveform more clearly.

Both input and output are sine wave with noise...the difference is the frequency, input is around 17.5 kHz and output is around 2 Hz

[Sam Chak]

Hey @CM Sahu

Why not plot the data exactly as you wish to present it? By attaching the image (click this icon ), we can better understand the input-output relationship, which may enable us to suggest a suitable model.

All data can theoretically be represented in various forms of knowledge-based models. However, it ultimately depends on how you wish to interpret the model to ensure it is easily understood by individuals within your field.

If you choose neural networks, then it will represent a highly sophisticated type of black-box model.

[CM Sahu]

input plot

input zoomed

output plot

Answer 1

[Shubham]

Hey Sahu,

I have tried plotting a subset of the data on the figure. Check out the following code snippet:

load('DATA.mat','data')
x = data(1:50,1)
y = data(1:50,2)

% Plot the original data and the fitted model
figure;
scatter(x, y, 'filled');
xlabel('x(t)');
ylabel('y(t)');

I get the plot as follows:

When I plot all the values, the data does not give much information, however based on the small subset of the data, I believe a high degree polynomial could fit this data.

Although it would be beneficial if you could provide more information on the data.

I hope this helps!

[CM Sahu]

Please read my above comments.

Plot for long data input you will able to see a sinusoidal wave like structure. column 1 is input and column 2 is output.... With frequency 17.5kHz and 2 Hz.

The model is acting like a frequency divider... I want a learning technique based solutions to map the input output relations.... Tried a lot but failed.... Please suggest..... Thank you in advance