Is this data noisy ?! Excel Dates in Plot problem too

[KealanOCarroll]

Level Control System, PID Controller I'll see can I tune any better.

I can fit straight lines to those plots really easily obviously, but is there any quick way to clean up the data or am I better off just leaving it as is?

E&H HART Pressure transducer at the bottom of the tank, 1m length of signal wire going through a backup power supply switching unit, into the ADC and profibus to the s7-300 PLC. The data stays as noisy when I stop the inlet flow so it's not sloshing in the tank as I first thought.

Also, WinCC Flex is outputting date & time in the CSV files, I've formatted them as both time and date in excel but it's putting the labels in a zeros along the X axis of the plot, anyone know is there a trick to sorting this?
The fact that both DD/MM/YYYY and HH:MM:SS are in the same cells seems to be throwing Excel

 

[danw]

What is the time base? It looks like 5 years of compressed data, from 1900 to 1905. What is it, really?

 

[iant]

can you paste the CSV file - you need to Zip it to do so

 

[KealanOCarroll]

XLS is All i have with me at the moment, that's attached if that's any good?

The actual time elapsed is 36 minutes

 

[iant]

I would like to see the raw CSV data - the 'time_ms' seems to be wrong the number is too large even for a 'REAL' number (428 days)

 

[Brian123]

Excel 2010 seems to be plotting the time series at the bottom of my chart just fine. I did have to select the TimeString column as the horizontal axis labels, but it is displaying them in the proper format.

As to the data, it does seem a bit noisy to me. Maybe add a filter in the PLC?

Brian

 

[Tom Jenkins]

Actually that doesn't look to bad to me. You can add filtering, but I'm not sure it will significantly improve control. It depends on the tank or sump, but some sloshing and fluctuation in level is not unexpected. (I am assuming that the formatting of the x-axis label is incorrect.)

 

[KealanOCarroll]

Got the excel bit sorted eventually.
After MUCH format painting and special pasting and such it turned out the X-Axis labels kept coming up as all zeros if there were two different columns of dates called.
The timestamps on the data are identical for each row of data, but unless the X-Values for both plots are selected from the same column it wouldn't display the dates.

On another note, I attemped to get the 'steady sine wave' Gain to work out some Z-N tunings. P control only. Upped the gain from 65 to 100; and got a response with no just like the above but with no oscillations around the SP ( I'm presuming because I turned off the integral action?).
There's anti-reset wind-up kicking in everywhere, which I expected, but the process time constant is a few minutes and the valve goes from 0-100 <20s so regardless of the gain It's never going to oscillate measurably about the SP.
What am I missing? I know there are processes with much longer time constants that use the method so I know I'm doing something wrong, I just cant find anything online about it, they just say use P-only and keep tweaking the gain & bump testing until the controlled variable sin wave-s around the setpoint. Then plug that Gain & Period into the tables to get a baseline tuning...If i'm understanding them properly :S

 

[shooter]

this is typical PID problem, you will see when opening mootrvalve it is constantly moving up and down.
slow down the I, switch of the D(the P will be enough)
and filter the input as it seems to have lots of interference from emc etc.
you can see the spikes clearly.
use a simple voltmeter.
switch of the HART protocol as many inputs can not live with the AC on the input