TECHNOLOGY Refinery experience provides guidelines for centrifugal pump selection

Gary R. Martin
Process Consulting Services Inc.
Grapevine, Tex.

Refinery pump services differ even when the properties of the pumped fluid are similar. For this reason, rote evaluation of pump specifications can result in pump reliability problems.

The selection and specification of centrifugal pumps for petroleum refinery service should take into account the inherent variability of the process service. Three case studies illustrate typical, chronic problems of pump reliability, evaluated from the process perspective.

Selection problems

Often, the process engineer does not fully understand process variability or effectively communicate it to the machinery specialist. In fact, most process engineers have only a rudimentary understanding of the complexity of a pump, let alone its seal system. As a result, they worry only about flow, head, net positive suction head (NPSH), and motor horsepower.

Selecting a pump is the responsibility of the machinery specialist. Communicating the inherent variability of the pump system, however, is the responsibility of the process engineer.

It should be remembered that proper pump selection and good service life are influenced by the process service, and all pump systems are not equal. Although pump specifications consider process conditions such as fluid temperature, vapor pressure, and flow rate, these properties do not reflect operating realities.

Often, the variation in flow of a given system is not well understood until a pump fails repeatedly. Pump specification criteria and vendor selection of the service should consider this variability. Arbitrary pump turndown specifications and selection of the lowest-cost pump can result in high pump maintenance costs and reliability problems.

Pump basics

The pump hydraulic system dictates the pump head-flow requirements. Fig. 1 [23172 bytes] shows examples of pump and system-head curves.

The pump curve is a function of the specific pump design. It is fixed once the mechanical design and pump speed are set. The system curve represents the pressure drop in the system.

A fixed-speed pump develops a fixed differential head for a given flow rate. The difference between the system curve and the pump curve is the variable pressure drop for the system. This value represents the action of the control valve. The flow or level controller sets the flow requirement and the valve moves to satisfy the pump head-flow requirement.

Pump specification sheets include data such as fluid specific gravity, NPSH, available pump-suction pressure, and pump discharge pressure. Using the data supplied on this sheet, the pump vendors select the appropriate pump for the service.

Occasionally, turndown requirements and more-specific data are provided in the specifications; more often, they are not. By default, process conditions such as turndown are assigned arbitrarily.

High turndown requirements affect pump selection. Typically, process engineers do not adequately report variations in fluid property for services such as interstage receiver pumps for wet-gas compressors on delayed cokers or fluid catalytic cracking (FCC) units. In these services, changes in fluid vapor pressure occur rapidly and the pump seal can be adversely affected.1

On many projects, the process heat and material balance is generated by one group of people and the equipment specifications by another. A rote approach to pump specification may work. But, to minimize reliability problems, refinery services require the input of someone that understands process variability.

Vendor pump specifications include efficiency, horsepower, and a curve showing NPSH required. Representative examples of these curves are shown in Fig. 2 [24280 bytes].

Pump efficiency is related to the specific mechanical design of the pump, its impeller size, and the pump speed. Horsepower is a function of pump efficiency, fluid viscosity, and fluid specific gravity. Refinery process engineers are familiar with these requirements.

Other pump design criteria-such as available NPSH, suction-specific speed, and pump power requirements at start-up speed-need to be considered.2 The dynamics of the process will place realistic limitations on the available NPSH. Suction speed is a function of the required NPSH, which is determined by the impeller suction-eye design.

The suction-specific speed is calculated using the following equation:

NSS = N (Q0.5/NPSH0.75)

In this equation, NSS is the suction-specific speed; Q is the flow, in gpm, at the best efficiency point (BEP); N is the pump speed, in rpm; and NPSH is the required NPSH at the BEP.

Pump service life is influenced by NPSH and suction-specific speed.1 The lower the required NPSH, the higher the suction-specific speed. In turn, the higher the suction-specific speed, the more likely suction-eye recirculation will occur. Under these circumstances, cavitation will be induced at flow rates less than the pump's BEP.1

While a pump with a high suction-specific speed can work well at the BEP, it may have a poor service life at 80% of BEP. In Case 1 (described in the following section) for example, a UDEX lean-solvent pump operating at 80% of BEP experienced erosion of the pump casing and impeller.

In recent years, refiners have begun to add specifications, such as maximum suction speeds, to the selection of centrifugal pumps. In some cases, the service turndown requirements are large; in others, turndown is limited by the operating limitations of critical equipment in the unit.

Refinery services include systems that cannot be reduced to less than 50% because of critical equipment limits imposed by the operation of other equipment. In other systems, the process flow can routinely approach 30% of design flow.

One example of the variability in some refinery services is the overflash pumps on vacuum columns. These pumps are subject to significant changes in flow rate. Although process operating objectives should address these changes, it is better to design the pump for the realities of the service rather than rebuild the pump on a quarterly basis.

For some services, there are process and instrumentation changes that can be used to mitigate large swings in pump flow and available NPSH. For others, the pump must be designed to accept the operating severity of the system.

All pumps have a minimum required NPSH and flow. These characteristics are inherent to the mechanical design, and operating below these minimums will result in pump damage.

Some pumps must handle a fluid that varies in composition from light hydrocarbon to water. The pump bearing and seals must be able to handle the imbalances created by this variation.

Solvent recovery

The first case history involves a bottoms pump system on the solvent recovery tower of an aromatics unit (Fig. 3 [52064 bytes]). The recovered lean solvent from the bottom of the column is used as feed to the top of the extractor.

This system comprised two pumps: one with a motor and one with a turbine driver. The motor driver was the main pump because the turbine discharge was 100 psig steam, which the refinery already produced in excess of process usage.

The motor-driven pump had a history of severe erosion of the impeller and casing. The carbon steel casing and impeller were replaced with a stainless steel case to eliminate casing erosion. This postponed replacement, but the casing life still was less than 18 months.

The turbine-driven pump had little casing damage. It operated at 80% of the speed of the motor-driven pump.

Service evaluation

The pump liquid was drawn off the bottom of the column. The reboiler was a once-through thermosyphon design withdrawn from a sump below the bottom tray. The reboiler outlet returned to the column below the draw-off.

The reboiler return piping design was such that the liquid level in the bottom of the column had to be kept low to avoid flooding the reboiler return. The liquid level was maintained within 2 ft of the reboiler piping return nozzle. The reboiler return was set at a low rate because the column was designed to maximize the number of trays.

The motor-driven lean solvent pump had very poor maintenance records. The pump mechanical seals, bearings, impeller, and casing needed periodic repair as a result of constant pump cavitation.

The selected pumps were double-suction, overhung pumps designed for an available NPSH of 12 ft. The low available NPSH resulted from the design of the column bottoms reboiler.

The selected pump had two available impeller designs. The impeller design with the low required NPSH was selected. This pump had a design flow of 80% of BEP. Fig. 4 [24903 bytes] shows the head-flow curve and required NPSH for the pump.

The pump head was about 300 ft. The pump suction-specific speed was calculated to be 17,600.

The cause of the problem was thought to be impeller- eye recirculation. When the unit was shut down, a test was run on the system to verify this theory.

The pump flow rate was maintained at operating conditions, and the column bottoms level was raised until cavitation stopped. The liquid level required to suppress the cavitation was almost 35 ft. The pump curve, however, indicated that 12 ft would be required.

An additional test was run. The flow rate was increased to a maximum. The horsepower of the pump motor limited the flow to 95% of BEP. At this flow rate, pump cavitation essentially stopped-as indicated by the discharge pressure stability, suction line vibration, and cavitation noise-even though the liquid level was only about 12 ft.

System variability

The lean solvent circulation was being operated at maximum rates to the downstream extractor. The limitation was hydraulic flooding of the extractor tray.

The tray design of an aromatics extractor will allow a maximum turndown of 50%. At rates less than this, the trays lose the dispersed-phase liquid seal, and solvent/oil contacting degrades. As a result, tray efficiency decreases markedly and aromatics recovery can deteriorate.

The pump selected for this service had a required operating range of 80-40% of the BEP flow rate.

Pumps with high suction-specific speeds are known to have turndown problems.1 To avoid cavitation, pumps with suction-specific speeds greater than 17,000 should operate above 90% BEP. In fact, this pump was field tested and the required NPSH was much higher than the pump manufacturer's curve.

The turbine-driven pump operated at 3,000 rpm. Pump efficiency declines as speed decreases, but so does the BEP flow rate.3 The combination of reduced speed and lower BEP flow rate decreased the pump suction-specific speed for the turbine-driven pump. The pump was operating at a higher percentage of BEP flow rate; hence, reliability was greater.

Interim solution

The refiner consulted the pump vendor. The vendor indicated that this pump design did not have a known history of maintenance problems. The vendor did, however, report that the pump had two available impeller designs. But the other impeller required more NPSH than was available. The higher-NPSH impeller had been designed much later than the original pump.

Capital constraints of projects have driven vessel skirt heights and available NPSH lower and lower. As a result, the reported pump NPSH curve for the impeller was incorrect. In fact, plant tests showed the required-NPSH curve looked more like a horseshoe than the typical curve shown in Fig. 2 [24280 bytes].

The long-term solution was to modify the reboiler piping to increase available NPSH and install the alternate impeller design with greater required NPSH. An interim solution, however, is illustrated in Fig. 5 [53855 bytes].

A pump spill-back (recycle around the pump) was installed. The spill-back allowed the pump to be operated at 95% of BEP flow, regardless of the required lean solvent flow rate in the extractor (Fig. 6 [24925 bytes]).

The recycle system installation cost 15% of the annual maintenance cost on the pump. As a result of the spill-back, annual pump maintenance costs decreased by more than 80%. The additional operating costs for the pump motor were minimal.

Vacuum overflash

Case 2 entails a refinery vacuum column and overflash draw pump (Fig. 7 [39816 bytes]). The overflash-liquid draw pumps are used either to recycle the material to the furnace or to meter the stream and return it to the bottom of the column.

Wetting of the packing in the wash zone of a vacuum column is very important; inadequate liquid leaving the packing will result in a coked grid bed.

The pumps selected for this service were oversized. Pump flow rates normally were less than 50% of the BEP flow rate. These pumps were subject to repeated cavitation because of problems controlling the liquid level.

Controlling the flow rate at a fixed value is an extremely important operating objective. In practice, the column liquid level is difficult to measure; therefore, the flow rate of this stream is difficult to control. Damage of the pump's mechanical seal as a result of cavitation is a chronic problem.

Service evaluation

All the liquid in this column is drawn from the collector tray. The level on the tray is measured by an external float system or a differential pressure measurement. The level controller resets the pump flow rate.

In principle, the flow rate from the collector tray should vary only when the column overflash changes. To operate at an economic optimum, the column overflash should be held at a minimum. Minimum overflash maximizes draw of the heaviest distillate.

In the field, the overflash liquid withdrawn from the column comprises both overflash and entrainment. The actual overflash liquid rate is a function of several variables, including crude charge rate, crude oil composition, transfer-line pressure drop, and furnace outlet temperature.

Overflash is difficult, or more often impossible, to calculate accurately using a steady-state computer simulation. Entrainment can only be estimated and is more difficult to predict than overflash.

Once the system is operating, the most important operating parameter is adequate wetting of the wash zone. A coked wash zone will shut down the unit.

The level-control system on this column is very unreliable. An obvious cause of level-control problems is plugged level taps. Neutron back-scatter level devices, however, also have shown foaming in the draw-off sump (revealed as changes in fluid density with elevation).

The oil may be cracking or the mixture of entrainment and residue may be generating vapor. The one obvious conclusion from these data, however, is that reliable measurements of liquid level are difficult or impossible to obtain.

A number of inaccuracies in the process design hinder the reliability of vacuum-unit overflash pumps. The design flow rate of the pump has to be estimated.

Typical systems for liquid level measurement do not work. The suction-line level of the pump chronically drops below the allowable minimum. Pumps in this service almost always are grossly oversized because calculating overflash is difficult.

System variability

The two sources of the feed liquid to the pump are true overflash and entrainment. During several plant tests in which the overflash was reduced essentially to zero flow, the overflash draw rate varied dramatically. The author believes the variation in flow is caused by changes in entrainment resulting from instability of two-phase flow from the furnace and transfer line.

The variations in overflash liquid flow are inherent to this service. Improved liquid-level control can help mitigate the problem. Nevertheless, the pump must be able to handle flow variations.

Variations in pump flow rate as low as 20% of the normal flow are typical and occur periodically in this unit. While this is not ideal, it is a reality of this system.

Solution

Flow rates in overflash pumps, except in very large vacuum columns, do not exceed 250 gpm and usually are less than 200 gpm. These pumps are small.

Available NPSH for this system usually is 50-70 ft. The solution to the cavitation problem requires fixing the level control and specifying a pump for maximum flexibility in flow rate.

Because most vacuum columns operate at very low pressure, the variation in column pressure has a negligible effect on the available NPSH. Column pressures generally are between 0.3 and 0.6 psia.

The suction pressure of the overflash pump is controlled 90-95% by static head; thus, the suction pressure more closely represents the static head of liquid.

The line diameter of the pump suction is larger from the column to about 15 ft below the draw-nozzle elevation. At this point, the line is swedged to an appropriate diameter for a pump suction line.

The larger suction line allows a slight increase in liquid residence time. It also allows gas venting back to the column. A closed-coupled pressure transmitter is used just above the pump suction. The liquid level in the suction line can be inferred from the pump suction pressure. Fig. 8 [47409 bytes] shows the proposed system.

The pump selected for this service should be designed for maximum turndown. It should have a suction-specific speed of 9,000 or less. Available NPSH, even when holding a liquid level in the suction line, is high. Pumps with low suction-specific speed are available for this type of service.

FCC compressor

Case 3 involves a compressor interstage separator drum for an FCC gas plant (Fig. 9 [53526 bytes]). The drum separates the cooled wet gas at an elevated pressure.

Generally, the pressure is about 75 psig. The condensed liquid is a mixture of hydrocarbon and water. Often only one pump is used.

The flow rate in this system is relatively stable. Variations in flow rate occur over a relatively long time (from day to night, for example) because of variations in the overhead condensation rate of the FCC main fractionator.

The liquid being moved by the pump comprises oil with a specific gravity of 0.65 and water with a specific gravity of 0.98. The composition of the liquid being moved by the pump varies between the water and hydrocarbon densities. This is revealed in the motor amp variations over time. These pumps have chronic seal problems.

Service evaluation

The available NPSH of the pump will vary as the fluid composition varies. When water is pumped, the fluid vapor pressure is 4 psia. When hydrocarbon is pumped, the fluid vapor pressure is 90 psia.

This variation presents a problem only when the fluid composition on the pump specification sheet is inappropriate. In this case, the fluid being moved through the pump is not of uniform composition.

Often, pump specifications will indicate fluid properties that are mixed (oil and water, for example), with a stipulation that the motor be sized assuming the total flow rate is water. Intermittently, the motor load will reflect the specific gravity of water.

For this service, the pump head requirement is about 600 ft, assuming a fluid with 0.65 specific gravity. The pump discharge pressure varies, depending on the specific gravity of the fluid being pumped. Not knowing the service, the variation in pump discharge may look like cavitation to many engineers. The pressure changes, however, because a pump puts out a fixed head, not a fixed differential pressure.

System variability

The pump flow rate varies with fluid temperature. The volume of condensation is strongly related to the temperature and pressure of the upstream column.

The flow rate of this stream varies, but is relatively easy to predict. The properties of the hydrocarbon, water, and mixed streams, also are easy to predict.

Solution

The problem with this pump is mechanical seal failure caused by vaporization in the seal. The light hydrocarbon flashes at atmospheric conditions. The fluid has an initial, true boiling point of -200 F. and a 50% boiling point of 65 F.

The seal design for this service is very important. Seal thermodynamics, while unknown to most refinery engineers, are conceptually simple.1 Seal technology, however, is somewhat complicated.

Noted machinery consultant Heinz Bloch recommends evaluating the seal design, then selecting a pump with the ability to accommodate the proper seal environment.

Acknowledgment

The author wishes to thank Heinz Bloch for his time and comments on the details of the mechanics of pump operation and reliability.

References

1. Bloch, H.P., Improving Machinery Reliability, 2nd ed., Gulf Publishing Co., Houston, 1982, pp. 83, 89, 85, 169, 172.

2. Karassik, I.J., Krutzsch, W.C., Fraser, W.H., and Messina, J.P., Pump Handbook, 2nd ed., McGraw-Hill Book Co., New York, 1986, p. 215.

3. Warring, H., Pumps: Selection, Systems and Application, 2nd ed., Gulf Publishing Co., Houston, 1984, p. 50.

The Author