Minimum flow for a centrifugal pump is supposed to ensure that liquid doesn't boil in the pump bowl. The refinery's standards dictated that pumps should have a minimum flow of 90% of the best efficiency point (BEP) flow — in this case, 180 gpm. That flow seems too conservative to me; by my rough estimate, it will cost the refinery $22,000/year.
Let's consider what pump minimum flow means and how you can ensure reliability without wasting money. At flows below the BEP, liquid will recirculate against the pump vanes, causing damage; you are doubly damned if the fluid is a slurry. The pump will become noisy, seal faces will heat up, and thrust bearings and shaft seals will experience increased stress. The minimum flow at which the pump will perform reliably is that just before the onset of cavitation, Qminc; cavitation is a chronic condition. The actual point at which this occurs varies with pump and application; the variables are: 1) suction specific speed — this is directly proportional to Qminc; 2) fluid properties — a lower density reduces Qminc; 3) continuous, or steady, pump rate; 4) the type of service — continuous or intermittent; 5) the margin between the net positive suction head available and the net positive suction head required — this varies inversely with Qminc; 6) impeller style — a decrease in diameter or increase in inlet vane angle reduces Qminc; and 7) the pump speed — this varies directly with Qminc. Finding the minimum acceptable flow is a complex problem that puzzles even experts. Pumps have spit out vanes after only a few weeks of continuous operation at 75% of BEP! On the other hand, I have seen pumps operated near deadhead for years without adverse effects.
As flow decreases below Qminc, the pump passes from the region of cavitation damage to a point, Qmint, that marks the onset of thermal damage, an acute condition that can cause almost immediate failure of pump seals; this condition is especially dangerous for canned and seal-less pumps.
For what they're worth, many rules of thumb exist for avoiding damage from recirculation. For instance, buy a pump with a continuous operating flow between 115% and 50% of BEP — 75% is ideal, according to www.pump-zone.com/topics/pumps/centrifugal-pumps/suction-specific-speed-part-two.
Now, let's consider the acute condition of Qmint. Another rule of thumb states that a 15°F-per-minute rise in temperature generally is safe for water; that's too much for hydrocarbons, though. The allowable temperature rise increases with pump size and decreases with heat capacity: ΔT ~ (42.4×BHPo)/(Wliq×Cpliq), where ΔT is in °F, BHPo is the shutoff pump draw in hp, Wliq is the pounds of liquid in the bowl, and Cpliq is the average heat capacity of the liquid. This equation is for a closed valve. For liquid flowing through a pump bowl use: ΔT ~ [(TDH)/(778×Cpliq)]×[(1-e)/e], where TDH is total dynamic head in foot of liquid and e is the pump hydraulic efficiency at a particular flow rate. If the fluid is mostly gas, "Cpliq" typically drops by a factor of ten and the temperature rise can be substantial. These equations are useful in checking Qmint, which is usually provided by the pump supplier. According to the "Handbook of Chemical Engineering Calculations," 3rd edition, Qmint in gpm can be estimated via: Qmint] = [4,668×BHPo×eo]/[(1-eo)×TDHo], where the subscript o denotes the value at the shutoff head.
Okay, how can you save money? Treat the acute risk separately with an intrinsically safe solution: design a parallel line with an orifice plate, or diaphragm valve for slurries, to meet Qmint. For the pump described above, the manufacturer set the flow at 30 gpm. Using the same head and efficiency as for the normal flow, because that's where the pump will operate continuously, the cost is $3,700/year — a savings of about $18,000/year.
For a low-risk intermittent pump or one with redundancy, consider a solenoid valve and orifice sized this time to satisfy Qminc. An automatic valve is more reliable than a modulating valve. Keep in mind that the valve probably requires a safety integrity level of two (SIL 2), i.e., a probability of failure on demand of 1/100–1/1,000; SIL 3 valves are available. Would you risk your entire operation at these odds? Obviously, a flow meter or flow switch works best for controlling the solenoid. Of course, process knowledge may permit a cheaper solution: a pressure or temperature switch or non-intrusive current coil, which would trip the motor off if the amps dropped; this solution depends on consistent fluid properties and an understanding of the system load and pump curve, and requires operator reset.
Coming up with an effective solution is easier if you can divide and conquer.