Outlier detection on measurement data

[MichaGue_01]

Hello everyone,

I am looking for a way to detect outliers of a sensor and delete them from my data. The challenge is, that I don't want to collect a number of data first, to recognize the outliers afterwards, but that with each new measured value it should be recognized whether it is an outlier or not. If it is an outlier, the value should not be displayed.

Attached you will find a vi with real measurement data which simulates the measurment in a for loop. In each cycle of the loop outliers should be recognized and deleted.

The picture shows an example what outliers are.

Any ideas how to go on?

Thanks a lot and best regards,

Michael

[GerdW]

Hi Michael,

a simple outlier filter according to your descriptions:

[MichaGue_01]

Hello, Gerd,

Thank you very much for your feedback and suggestion. What I forgot to say is that the interesting measurement value can also fluctuate (see picture, red arrows mark the outliers) and is not constant as in the first example. In principle, the process can look like this, i.e. the limits to recognize the outlier have to adapt automatically.

How could one proceed here?

Thank you.

Many greetings
Michael

[GerdW · Accepted solution]

Hi Michael,

now it is getting trickier…

What about a (Pt-By-Pt) Median filter?

[MichaGue_01]

Hello Gerd, thank you very much. You gave me the right suggestion. Super!