Hi everyone, please ask a question. How to remove bad values and filter the experimental data.The data is shown below.

[Chris]

Okay, I'll ask a question. How do you define a "bad value"?

[Wesley]

Thank you very much for your kind reply. My definition of bad value is to remove the high-frequency part of the data and get the low-frequency data.

[Chris]

Okay, there is actually a bad value in this vector. You can find it with

badValues = ~isfinite(x);

Then you could choose whether to delete it, or replace it with a 0 or an averaged value based on neighboring values.

After that, you should be able to apply a low-pass filter to quiet the high-frequency noise if you have the Signal Processing Toolbox, or smoothdata, or something along those lines.

[Image Analyst]

I already repaired the data. Don't miss my answer below.

Answer 1

[Image Analyst]

Try movmean():

s = load('24v.mat')
u = s.u;
x = 1 : length(u);
% Interpolate nans
nanLocations = isnan(u);
u = interp1(x(~nanLocations), u(~nanLocations), x)
% Now u has been "repaired" by filling in nans.
% Plot repaired data.
plot(x, u, 'b-')
grid on;
% Smooth with movmean() function.
smoothu = movmean(u, 201);
% Plot smoothed data.
hold on;
plot(smoothu, 'r-', 'LineWidth', 4)
xlabel('Index', 'FontSize',17);
ylabel('Signal', 'FontSize',17);
legend('Original Signal', 'Smoothed Signal')

[Wesley]

Thank you for your enthusiastic answer, it is very helpful.