Recording Amps on two screw compressors at different sampling rates

[compguy22]

Hello,

I am recording Amps on motor's which are driving rotary screw compressors. Total is two compressor units, each of which is discharging into a common header. Amps are recorded at motor's electrical supply ; the objective is to post calculate, starting from the Amps, each unit flowrate profile over time. Subsequently, the sum of the two flowrates will allow to establish how the total flow demand fluctuates (feeding downstream users).

Problem is that the two units are not sampling/logging Amps measurements at same rate, thus the sampling time stamps are out of synch. For some reasons, sampling is not at regular time intervals, so data are irregular scattered.

One option I am evaluating is to use the machine with lowest sampling rate (call it machine A) to obtain reference sampling times or time steps. So the lowest resolution will be the driver or the limiting factor on the total resolution. These reference sampling times will not match one to one with the second unit (machine B) with highest sampling rate. To try to construct a match, I thought I could use linear interpolation on the second machine profile (for example, use 1D linear regression based on irregular scattered data).

Do you think this can be a sound approach? Is there any option that would more straightforward?
Somehow this links to theory of information or not? Should I post in another section of the forum?

Thank you

 

[compguy22]

For reference, I found this article on so called cross-correlation, all though I do not understand it exactly yet:

https://uk.mathworks.com/help/signal/ug/align-signals-using-cross-correlation.html

 

[SwinnyGG]

Putting a pressure sensor in the header between the compressor outputs and whatever valve system holds pressure would save you a bunch of characterization math and eliminate having to sync two signals. Just saying.

 

[compguy22]

@SwinnyGG thank you.
Could you please elaborate on the set up that would help establish flow demand profile from pressure measurements?

 

[SwinnyGG]

If you know pressure and the dimensions of the plumbing containing that pressure, then you know flow.

 

[compguy22]

So basically you would be measuring the pressure downstream say a piece of pipe, after the compressors; knowing the outlet pressure of the compressor you can determine pressure drop and then from pressure drop you then calculate flow rate based on pipe length and diameter. Is this correct?

 

[SwinnyGG]

yes.

 

[compguy22]

what if compressor outlet pressure is fluctuating, for instance due to load and unload cycles ? how about the case of variable speed compressors or even piston type units? Could you please share your experience.

 

[SwinnyGG]

Pressure changes are captured by.... wait for it.... pressure sensors.

 

[compguy22]

If I have access to controller or plc data, then I may even have the flow value right away, not even need to do measurements.
The problem is that not all units have gateway communication in the compressor room; so pressure changes are captured by pressure sensors but if I cannot immediately access the logged data then what is the benefit. Many old units are not equipped with gateway communication or have basic display panel. Read out from indicators and pressure gauge are most of the time available, but here we are considering a continuous recording at regular and fast sampling periods.

By the way, as regards installing a pressure sensor in the system downstream, be it in the common header or elsewhere, sometimes it requires hot taping and this can be complicated or even subject to permit.

 

[SwinnyGG]

At some point you’re going to have to engineer something. We can’t do it for you.

 

[compguy22]

No disrespect but I am not asking for that.
I refer you to my OP, the question is basically about sharing experience or any hints on how to go about processing two signals / log data that are out of sync. I used linear interpolation, I compared down sampling and up sampling, I used LOOKUP function, etc.
Somehow I find a way around it, but when someone share hands on experience, when possible, it is advisable.

 

[SwinnyGG]

I (and many others) have recorded, characterized, or otherwise utilized process variable data taken from native equipment PLC sources, or from separate data acquisition systems installed for this sole purpose; ultimately the right solution to the incomplete problem you've described depends on a bunch of factors you haven't described in any way.

If your on board data sources don't have enough fidelity, you will need to add data sources. If you can't do that, you'll need to make due with what is already available to you. That is the best information that I, or anyone else, can provide based on what you've described so far.

 

[compguy22]

'...you've described depends on a bunch of factors you haven't described in any way.'

When I am asked specific questions by you the 'expert', then I could try provide specific answers, then I could improve my understanding of what factors come into play and what I am missing in the picture. This interaction is valuable in the process of problem solving. And, if I do not have the answers or know the details, I can at least try to look for it or ask around me. That is what I am expecting from triggering a discussion thread. Now I feel I have to ask you what is complicated about it?