Plant testing and troubleshooting often require data gathering to identify specific problems. Data gathering needs vary with the process and the equipment. Data must reflect fundamentals of the equipment operation. Getting usable results, of course, is crucial — but data gathering efforts must contend with time and money limitations as well as safety constraints. Wonderful high-precision instruments are available. However, they may be too difficult to use or too expensive for many routine tasks. Troubleshooters must know the limits of conventional measurement instruments and understand what data are necessary to come to usable conclusions. An examination of centrifugal pump performance illustrates the need to adequately appreciate instrument capabilities.

[pullquote]

Every centrifugal pump has a performance curve. As flow rate increases, dynamic head across the pump drops. So, knowing dynamic head across a centrifugal pump can give information about flow rates. The key question then becomes how accurate must the required pressure readings be to estimate dynamic head.

Figure 1 highlights the challenges posed by pressure gauge accuracy on estimating flow rates. This particular pump was installed in a liquid recovery system that collected slop streams. The nominal design rate was 50 gpm but normal “expected” operation was 35 gpm. The system did not meter flow. One part of the troubleshooting effort required getting an accurate estimate of the liquid rate. This pump has a constantly rising dynamic head down to shutoff (zero flow). One characteristic of a pump with a constantly rising head is that the dynamic head changes only a small amount with flow rate as that rate approaches zero.

**Pump Curve**

Figure 1. A modest inaccuracy in pressure measurement can translate into a far greater error in flow rate.

The suction pressure was 15 psig. To get a dynamic head reading, the troubleshooters used two pressure gauges: one on the suction and one on the discharge. The maximum expected discharge pressure was ~90 psig based on a 0.7 specific gravity fluid. For safety, an initial check was done with a 0–200-psig gauge on the discharge, then a 0–100-psig gauge was used for the final readings. Because no lower-pressure-range gauges were available, a 0–100-psig gauge was used for the suction pressure as well.

If the pump were operating at the normal rate of 35 gpm and on the pump curve, the gauge on the discharge would give a reading of about 82 psig. The best case for standard commercial pressure gauges found at most plants is an accuracy of ±2% of the range for readings. So, the 0–100-psig gauges used for the suction and discharge pressure readings each should have an error of ±2 psig. Additionally, both gauges will be taking readings close to the ends of their range: the 15-psig suction pressure is in the low end of the range while the 82-psig discharge pressure is in the high end. Readings like these towards the end of the range likely will have higher errors than readings more in the middle of the gauge’s range.

To get a number, let’s examine the effect of a 2-psi error in the dynamic head. If the actual pressure is 82 psi but the gauge reads 80 psi, the height is 215 ft and the estimated flow rate is 44 gpm. If the gauge instead reads 84 psi, the height is 228 ft and the estimated flow rate is 22 gpm. The flow error range is -36% to +25% for a reasonable error in measuring dynamic pressure. Lower flow rates shift all readings to the right on the curve, making likely errors greater. Higher flow rates shift the readings to the left on the curve, making flow errors smaller.

The accuracy needed on the flow measurements depends upon the problem you are trying to solve. However, troubleshooting rarely benefits from measurements with a -36% to +25% error range on flow rates. At high enough rates, pump discharge pressure might give a useful indication of flow rate but it’s never the best method.

A better option for pumps with motors is to get an estimate of pump power from the electric load and then calculate a flow rate. In this case, the flow-rate error drops to -18% to +10%. While this certainly isn’t great, it still is a 50% improvement.

Of course, these analyses all depend upon knowing the pump curve and assuming the pump accurately follows it. At low flow rates, small deviations due to wear or other factors will significantly change any flow rate estimate from pump or motor data.

The case discussed is a difficult one. The best solution here, if possible, is to directly measure flow with an ultrasonic flow meter. Nevertheless, the case dramatically illustrates the importance of understanding the accuracy of measurements and how they may affect troubleshooting.