Ask The Experts: Controlling levels of parallel separators

This column is moderated by Béla Lipták, automation and safety consultant and editor of the Instrument and Automation Engineers’ Handbook (IAEH). If you have an automation-related question for this column, write to

We have a horizontal separator that works under level control (with gas-blanketing on top) and feeds a set of parallel-operating pumps. The level control valve (LCV) is at the pump discharge header, and gets the control signal from the level transmitter (LT) on the separator.

To handle increased influent flow rate to the existing separator, a new separator roughly half the design capacity of the existing one will be added to work in parallel. Two new pumps will be added to handle the additional effluent from the new vessel.

The separators are for separating oil from water. The exact size of the new separator is unknown for now as it is yet to be sized. Operations people want the new separator to have the same diameter and to operate at the same normal liquid level (NLL) as the existing separator, so the liquid level in both vessels could be controlled with one level control valve at the pump discharge. The operators’ wish is not the designer’s comman d, of course, as the separator dimensions and NLL are dictated by residence time considerations, and fixing it because of level control considerations is perhaps not a step in the right direction.

I’d be obliged if you could review the three sketches A,B and C (Figure 1) and comment on how to best control the levels in the two vessels.

Farooq Ghilazi /



Option A will work if the inlet flow distribution is guaranteed by hydraulic design, and is the easiest to implement. Another option is described below, which will also work, but is not necessarily superior.

The key to good separator control is keeping residence time (volume/flow) above the required minimum for good separation. In addition, good blanketing gas, water interface level control and oil level control are required. For the purposes of this discussion, the gas pressure and the interface controls are assumed to be properly designed.

When controlling two separators in parallel, distribution must also be controlled. In Figure 2, I assumed that the distribution is 2:1 and is taken care of by hydraulic design. If it is not the case, a flow ratio control loop must be added. This addition naturally not only increases the first and maintenance costs, but also, because it increases turbulance, increases the residence time required and therefore lowers separation capacity.

As to oil-level control, one should equalize the oil levels in the two separators by desiging a balancing pipe that is hydraulically capable of balancing the levels, and can also be used as the pump station’s suction manifold. In that case, when both separators are in operation, the level measurement is obtained by averaging the two transmitter signals. When operating only one separator, this averaging function is bypassed.

As to the pump capacity controls, I would not waste energy by valve throtting on the discharge side of the pumps. Instead, my preference is to have at least one variable-speed pump in the station, so that pumping capacity can be modulated without energy waste.

For a good book on hydraulic design for residence time determination, refer to

Béla Lipták /



It seems to me that the residence time in one separator can’t be modeled after the time in another. It might work if both separators are identical and cleaned frequently. It also might work if an operator periodically, physically checks the interface levels in both separators. I don’t know what he’d do to fix an imbalance—possibly stop a pump for the low tank. Operations won’t like that.

Maybe it would work if the inlet flows were controlled according to the size difference. If you tell operations that they have to add two flow control loops to save one level loop you would get the correct result, which is separate level control loops. Also, throttling inlet flow could further mix oil and water, increasing the separation/residence time, which increases the tank size for a given flow.

White Paper: Wireless Connectivity for the Internet of Things

Maybe it would work if one of the new pumps had variable-speed control, which is set by a separate level controller in the new tank. You’d need logic to start the other pump if the variable pump got to 80% of full speed or so. That saves the cost of a new valve and its installation and maintenance.

Disclaimer: I’ve designed and started up controls for many things, but not separators.

Bill Hawkins, HLQ Ltd. /



Figure 3 shows a simplified sketch of how you may be able to connect the two separators to a common inlet and outlet header. The level indicating controller (LIC) will control the valve in such a way that the highest level is pumped down, and with the High-Low selector, you can automatically or manually select any of the separators. This way, if the level is too low, you can stop pumping.

This, to say the least, is a simplified sketch. For detailed loop design, more data needs to be analyzed.

Alex (Alejandro) Varga /



You’re correct to be concerned about how the tanks are tied together. One way I can see to possibly make it work would be to have a separate, “large enough” balance line between the tanks, but that would require modifying the existing tank.

Otherwise, what the individual levels end up being depends on the hydraulic balance on the inlet piping and the outlet piping. You can end up with one tank on high level or one running empty.

Another possible solution would be to rearrange the piping, so you have a larger, common suction header to the pumps. This will mean that it can act as a sort of balance line, and then you would run the control on high-level override, with the valve controlled by whichever level is the highest. I’m not sure this will work; it’s a question for the process engineers to confirm. But the real solution is not in the controls; it has to be a workable hydraulic design.

Simon Lucchini, Chief Controls Specialist /