Data smoothening

[Ali768]

Hello

I have done some experimental research work and found out from data a kink (in the middle of the data )at the switching of the two flow meter lines (while using air)i.e. when i switch from low flowline to high flowline via controller. Obviously this is because of the low flowmeter is at highest value while the highflowrate meter is at the lowest so the uncertainity level is high at these region.

Now the problem is that I have use this data to calculate various other parameters and using this kink data results in kink in all paramters.

I was wondering if somehow i could smoothed the data so that kink of the flowmeters data is smoothened and does not appear on the various other paramters calculated.

Any suggestions/help will be appreciated

Aly

 

[IRstuff]

Why do you need to calculate anything during a transitory condition?

TTFN

 

[Ali768]

It is not trasitory condition ............ I am using air flow through the pipe and that is used somewhere in the experimental setup ........... because i need a wide range of air flowrates, I had to use a existing compressor setup which uses two flowmeter lines one for low air flowrate and other for high flowrate. Low flow meter is working at it highest value i.e. at 150Sm3/hr while high flow meter at the minmum value of it range i.e. 200 Sm3/hr. my data points consists of 50,100,150,200,250,...500Sm3/hr. The kink mentioned above is coming at 200Sm3/hr where the smooth curve from (50,100 and 150) data shifts upward by a offset at 200 and then again smoothly follows the trend till 500.

I hope i have cleared your question.

Aly

 

[IRstuff]

Not really, but, why not just do a running average? or a median filter? or any number of low-pass filters?

TTFN

 

[SomptingGuy]

Spikes, kinks and other time-domain phenomena aren't easily filtered out using classical filters, because they have a broad spectrum. Low pass filtering (with a moving average or other such filter) will smear the spike/kink but not remove it.

If there is just one kink and you know where it is, can you not stich the data together by hand? You are working in Matlab and that's a godd environment for this sort of activity.

 

[MikeyP]

Maybe the best approach is to leave the data as it is and include an explanation as to the cause of the kink. It is better to be honest rather than to adjust the data without any justification.

M

--
Dr Michael F Platten