We have a troublesome system, it goes like this.
A centrifugal pump has a control valve in the discharge, downstream there is
a header that fans out to 5 offtakes, each of them has a flow control on
them with an individual valve. The common valve off the pump is driven by a
pressure controller to maintain a constant pressure downstream, ie. upstream
of all the offtakes. All the controllers are standard industry PID, scan
time 1 second.
The system cycles intermittently, but most of the time, at a frequency of
around a cycle a minute, with an amplitude cin the pressure that is about
10 - 15% of transmitter span. That's large enough ot be a real problem here.
All of the control valves are butterfly valves, and are running at a
sensible point in their range, ie. around 30% open. There are no significant
other lags in the instrumentaiton, all the flow transmitters have a time
constant of less than a couple of seconds. From what I can see, none of the
valves suffer from significant mechanical imperfection - ie. greater than a
percent or so. The pump is well sized, it's not running on a steep part of
its curve. Inshort, there don't appear to be any obvious problems with the
setup. Pipes are big, hydraulic loss is minimal, and there's no significant
static head involved.
Downstream of the flow control valves is a reasonably steady pressure, there
doesn't appear to be any external periodic disturbance present. When the
pressure controller is set to manual, ie. valve position fixed, the
oscillation stops. If all the controllers are active, then the tuning
settings need to be set to very (unacceptably) loose values to make the
oscillation go away. A dynamic simulation with reasonably accurate
representation of everything in the loop doesn't display the instability we
Being a fairly heavy duty industrial setup involving non-benign fluids,
intervention such as putting on additional measurements or making trial
changes is a no-go. Frankly I'm stumped, has anyone come across this
configuration anywhere that shows an unexpectedly high tendency to limit
10 years ago