Nearly every process input is a flow, whether directly via a material input flow or indirectly via an energy input (e.g., utility flow). Good flow control is important for achieving the desired material and energy balance and stoichiometric ratio for reaction rates. Flow control depends on meters with good precision and turndown and upon tight pressure control. Liquid pressure and flow loops have the same fast process time constant that is typically less than 1 second. The main game for tight pressure control is to minimize deadtime and not introduce any delusional time constants via signal filters. Here are the key considerations.
This discussion and checklist assumes liquid control valves with a vena contractor pressure greater than the liquid vapor pressure so there is no flashing and flow changes are a result of either a flow control valve movement or a change in the valve inlet or outlet pressure (valve pressure drop). Fluid density can also change the flow but typically density changes are much slower and are best corrected by a mass flow meter. We are addressing fast disturbances.
The interaction between a liquid pressure controller and flow controller in series can be minimized by making the pressure loop faster and designing the pressure drop across the flow control valve to be much larger than the pressure drop across pressure control valve. Normally we think of the pressure control valve upstream of the flow control. This is the case for header pressure control with several flow control loops off of the header. However, for a single flow loop and pressure loop in series in a pipeline, the pressure control valve could be downstream of the flow valve, giving us an option for pairing control valves in series with the pressure and flow loops.
A fast pressure loop that can minimize the peak error and integrated error for pressure disturbances will minimize the impact on the pressure at the flow valve inlet or outlet. Consequently the total pressure loop deadtime should be much less than the total flow loop deadtime. Also, the pressure controller must be tuned aggressively to reap the benefit of the deadtime advantage. This involves as much gain action as possible. The maximum controller gain is the largest time constant divided by the product of the deadtime and open gain. The open loop gain is the product of the final control element gain, process gain, and measurement gain. The final control element gain is the slope of the installed characteristic of the control valve or variable speed pump (flow change per % signal change). The process gain is twice the pressure drop across the pressure control valve divided by the flow. For a low valve pressure drop to flow ratio, the process gain is low. The measurement gain is 100% divided by the measurement span. For a large measurement span relative to the operating range, the measurement gain is low. A low deadtime, low process gain, or low measurement gain translates to a large maximum controller gain. Reiterating, the pressure loop PID should be tuned with the maximum possible gain.
If a measurement filter is added that is much larger than the process time constant, the controller gain can be increased but what you see in the trend chart is a filtered pressure. The real (unfiltered) pressure will generally be worse. You are fooling the controller and yourself.
An equal percent characteristic has a slope that is proportional to flow that will help compensate for the pressure process gain nonlinearity that is inversely proportional to flow. Since the pressure control valve drop is by design small compared to the system pressure drop, the installed characteristic is close to the inherent characteristic, and the compensation is accurate.
Many of the ways mentioned for minimizing deadtime for gas pressure control in the last checklist are even more important for liquid pressure control because the process time constant, an inertial lag, is less than one second for non-compressible flow unless we are controlling pressure in a very long header.
I speculate viscous drag may slow down the start of flow through a control valve much like trying to get ketchup out of a bottle. I have not seen the first principle equations to quantify this effect but I have seen it in terms of a slow start of viscous 98% sulfuric acid reagent flow for pH control.
At any rate we are talking about very fast dynamics. In automatic, the trend chart excursions are so fast, they may look like noise. I don't advise putting the loop in manual, but if you did you would see these are real disturbances that are being attenuated by pressure control. Tuning cannot generally be safely done based on open loop responses.
A self-contained field pressure regulator is essentially a high gain proportional only controller with no measurement, communication, execution, or actuation lag. While a pressure regulator can potentially provide tighter control, the lack of visibility and adjustability makes a pressure loop desirable. The pressure loop PID execution time must be 0.1 seconds or less, the transmitter damping 0.1 seconds or less, and the time to respond to the change in controller output needed to reject the largest disturbance less than 0.1 seconds. For tight liquid pressure control a pneumatic actuated control valve is not fast enough. A hydraulic actuated valve or variable frequency drive with negligible deadband or rate limiting must be used. In cases where the final control element dynamics are negligible and disturbances are particularly large and fast, a PID execution time of 0.025 seconds or analog controller may be needed.
Direct mounting of pressure transmitters is preferred for pressure control. Diaphragm seals and capillary systems and wireless transmitters are too slow for tight liquid pressure control.
To minimize interaction between loops, relative gains close to one are desirable. The relative gain for a flow loop in series with a pressure loop is proportional to the ratio of the flow control valve pressure drop to the total system drop. This is consistent with the statement that the pressure control valve drop should be much smaller than the flow control valve pressure drop. For flow loops in parallel, the relative gains of the flow loops approach one as the number of flow loops increase. For a relatively flat pump head curve, the relative gains also approach one. See the 4th edition of Shinskey's Process Control Systems pages 248-251 for more details on the relative gains for liquid pressure and flow control.
The flow process gain is one. The flow measurement gain is linear (100% divided by flow span). If the flow final control element gain is linear, we have the rare case of a linear loop. A VFD pump on a system with negligible static head has a linear installed characteristic. An equal percent characteristic where the control valve drop is a large fraction of the system drop has nearly a linear installed characteristic. Thus, there is a double benefit of allocating a large pressure drop to the flow valve in terms of minimizing interaction with the pressure loop and linearization of the flow loop.
The flow loop is generally tuned less aggressively than the pressure loop with more integral than gain action. Also, the flow loop deadtime is not as critical and can be 1 second or more. Thus the PID execution time can be 0.5 seconds and the control valve response time can about 2 seconds.
A severe limitation for flow control is flow measurement turndown. The lowest flow that can be controlled is the flow where erratic behavior or noise is less than the allowable error about setpoint. For differential head measurements with extended range or dual range differential pressure transmitters, the turndown from maximum scale flow is about 8:1. For vortex meters the turndown can be 12:1 if the maximum scale flow corresponds to the maximum velocity for the meter size (a rare case). Differential head and vortex meter turndown assume a good velocity profile from sufficient straight run of a well designed upstream piping system.
Many of the considerations mentioned here also apply for pipeline gas pressure and flow control loops in terms of pressure drop allocation and aggressive tight pressure control. Gas pressure control on columns and vessels benefits from a process time constant that is orders of magnitude larger enabling less stringent requirements on the pressure loop deadtime. Furnace pressure control is another story.
Here is the checklist to help ask yourself the right questions for tight liquid pressure and flow control. For less demanding liquid pressure and flow control loops, the cited times can be a factor of 2 larger.
•(1) Is the PID execution time 0.1 seconds or less for liquid pressure control and 0.5 seconds or less for flow control?
•(2) Is the total of the measurement sensor lag, transmitter damping setting, and filter setting, less than 0.1 seconds for liquid pressure control and 0.5 seconds for flow control?
•(3) Does the pressure loop final control element need to be a hydraulic actuated control valve or variable frequency drive with negligible deadband and response time?
•(4) Was the pressure control loop tuned for maximum controller gain?
•(5) Does the pressure PID need to stay in automatic to prevent an excessive pressure excursion?
•(6) Is interaction minimized by making the flow control valve drop much greater than the pressure control valve drop or operating on the flat part of a pump curve?
•(7) Does the pressure control valve have an equal percent flow characteristic to compensate for the nonlinearity of the pressure process gain?
•(8) Does the flow control valve have an equal percent flow characteristic and is the flow valve drop a high fraction of the total system pressure drop to help keep the flow loop linear?
•(9) Does the flow meter have enough turndown for the selected meter size and calibration?
•(10) Is the flow control valve resolution limit better than 0.1% and deadband less than 0.2%?