Emerson Exchange 365

• Not Answered

RAMP detection, for anticipation and prevention of pressure drop in ethylene header by start of hyper compressor in LDPE plant.

Ethylene Header has a main pressure controller which keep the pressure in 40.5 Kg/cm2 , but one of our consumers is a LDPE plant which in start up steps causes pressure drop in header because of hyper compressor. 

I would like to implement a ramp detection in order to detect o anticipate to this pressure drop. Does anyone implement something like this? Best Regards!

jtax

• 7 Jun 2018 1:44 PM
• Cancel

4 Replies

• James Beall

  

  • 14 Jun 2018 9:03 PM

  I am not sure if anyone does this but it is certainly possible to configure a DeltaV module to calculate the rate of change of the header pressure to detect a step change in the header load due to the compressor start. One method is input the header pressure to a deadtime block and subtract the OUT of the dead time block from the current header pressure and divide by the deadtime setting of the deadtime block. This will be the rate of change ((Pressure X second ago - Current Pressure)/(X seconds)). This signal can be used in the control scheme. I would need to know more details about the system to make any additional comments. Hope this helps!

  • Cancel
• Rafael Rebouças de Araujo

  

  • 15 Jun 2018 10:04 AM

  Be very careful with measurement noise, you will have to implement something to take care of it.
  If you do need some logic to detect these kind of pressure drop, probably the hardest part will be to make sure that the logic will not be triggered spuriously, nor will be prevented to act (or act late) when needed (due to excessive filtering for example).
  
  Also, no need to say that I hope it is not a huge and important action. If you are betting everything on this kind of logic, maybe something is wrong with the design or somewhere else.
  
  It's hard to say without looking the data and doing some tests, but if the pressure drop isn't too fast you could use for example least squares to estimate the slope. Other option would be using a standard ROC block on delta v and try filtering to check if the results are acceptable.
  
  You should also let this logic online (making the calculations) but keep it disabled (prevented from acting) until the pressure is below some value, to prevent acting when you have a drop above setpoint. You could do something like:
  
  If drop is bigger than XX and pressure is less than YY
  Trigger actions
  
  Well, actually I think you will bump in many obstacles that you will only realize when making the tests. I already used least squares to calculate a level slope to estimate the flow in/out a vessel, it works well depending on what's the final objective. I never did on Delta V, but it's pretty straightforward and I can say that Delta V has many more tools than the system in which I did it.

  • Cancel
• Luis Navas

  

  • 15 Jun 2018 12:54 PM

  Hi, If I understood properly….one option could be the use of arrays, storing on it the PV and after comparing after an x time, the current pv vs an old pv inside of array,( keep in mind the module scan rate).

   

  Regards,

   

  Att. Luis Navas

   

   

   

  • Cancel
• Rafael Rebouças de Araujo

  

  • 15 Jun 2018 8:02 PM

  Simple question: what is your pressure controller doing? Admitting more/less ethylene into the header? A simple tuning or maybe adaptive tuning wouldn't be enough?
  Did you consider a FF strategy?

  • Cancel