trend analysis

[dave56]

Are there any VI's that allow me to do trend analysis? I have a graph of data and would like to able to have an alarm or something alert me to aberrations. This could be from a drop off in recorded data or any unusually high or low readings.

My data can be erratic sometimes so simply just taking the mean and alerting me to anything that is X units above the mean won't really work because an occasional reading will spike so I would like to be able to set off on sustained spikes or valleys.

[JuanCarlos]

Hi,

The specifics on the functions that you use will depend on the trends that you want to detect. You could use a low-pass filter to get rid of spikes or noise and then compare the level. You can also use a liner fit to see if the data has a trend upwards or downwards.

If the data is coming in one sample at a time, you can use the point-by-point version of the analysis VIs; they will keep the buffer for you and do the running processing.

I hope this helps. Let me know if you have any further questions.

Regards,

Juan Carlos
N.I.

[dave56]

The data isnt real time or anything, it is read in from files collected.

I think the objective in this analysis would be more about seeing if there is a sustained aberration from the norm, as opposed to just several points spiking above or below the median.

I am not too sure about which Analyze vi's would be most appropriate to use however.

[JuanCarlos]

Have you tested the Histogram VI's. you get a good idea of where the values fall given a set of points. You can also perform a shot time histogram analysis to get a dynamic view of the trend. Once you have the data separated in chunks you could so some sort of running average and standard deviation to see how accurate the point is.


Here is a small example that shows the use of a histogram, mean and standard deviation. You can see how the spikes are smoothed in the histogram and how the standard deviation can indicate where a spike occurred.

Let me know how this goes.

Regards,

Juan Carlos
N.I.