HKIE Professional Assessment - Control, Automation & Instrumentation Discipline: Steam flow reads high at part load - impulse line condensate or a worn orifice plate?

Steam flow reads high at part load - impulse line condensate or a worn orifice plate?

At a food plant, saturated steam at 1.0 MPa(g) is measured by an orifice plate with a DP transmitter ranged 0-100 kPa, which the flow computer square-roots and scales to 0-20 t/h. Both impulse lines run to condensate pots and the transmitter sits 600 mm below the taps. Since the last annual shutdown eight months ago, the steam flow indicated at medium load has read about 5 % higher than the plant heat balance derived from condensate and feedwater metering, while at full load the two agree to within about 1 %. The transmitter was calibrated on the bench three months ago and its zero was checked with the three-valve manifold equalised; it read 0.2 % of span. Line pressure is steady at 1.0 MPa(g) and ambient temperature varies by less than 10 K. The instrument technician proposes to apply a 5 % bias in the flow computer, arguing that the orifice plate is probably worn. The energy manager wants a defensible answer because the steam bill is allocated across three production lines.

1. Calculate the indicated flow error at 25 %, 50 % and 100 % of maximum flow that would result from the low-pressure impulse line condensate level being 250 mm lower than the high-pressure line, and say whether this explains the reported observations.

Read the worked answer

Physics. The transmitter sees the true DP plus the difference in static head between the two condensate columns. If the low-pressure pot level falls (for example a weeping drain valve or partial flashing), the LP static head falls and the measured DP rises by 250 mm of condensate.

Head: 0.250 m x 900 kg/m3 x 9.81 = about 2.2 kPa (condensate near saturation temperature, density roughly 900 kg/m3). That is 2.2 % of the 100 kPa span - an additive offset.

Indicated flow equals full-scale x square root of (measured DP / true DP):

25 % flow: true DP = 6.25 kPa, measured 8.45 kPa, ratio 1.352, root 1.163, indicated 29.1 % - +16 % error
50 % flow: true DP = 25.0 kPa, measured 27.2 kPa, ratio 1.088, root 1.043, indicated 52.2 % - +4.3 % error
100 % flow: true DP = 100 kPa, measured 102.2 kPa, root 1.011 - +1.1 % error

Does it fit? Yes. An additive DP offset produces a load-dependent error that is worst at low flow and almost disappears at full flow, which is exactly the reported pattern (about 5 % at medium load, about 1 % at full load, and much worse at low load - which the plant would see as implausible overnight flow readings). A 250-300 mm level difference accounts for the observed 5 % within measurement uncertainty.

Rule out the technician's hypothesis. A worn orifice edge increases the bore, which reduces DP for the same flow, so the meter would read low, not high. The same applies to a reversed plate. A 5 % bias would also be wrong everywhere except at one operating point: it would leave the 16 % error at low load uncorrected and would over-correct at full load, and it would conceal a fault that can grow.

2. List the other plausible causes of a high reading and describe a test that would confirm or exclude each. Explain why the bench calibration and the equalised-manifold zero check do not exclude them.

Read the worked answer

Static pressure effect in the DP transmitter. Typically around 0.05 % of span per 0.7 MPa. At a steady 1.0 MPa(g) this is well under 0.1 %, so it cannot produce 5 %. Check the vendor specification and confirm the line pressure is genuinely steady on a 24-hour trend.
Ambient temperature effect. Around 0.1 % of span per 10 K. The stated swing is under 10 K, so this is negligible. Confirm with a cabinet temperature log.
A leak in the high-pressure pot or line. This gives an offset of the opposite sign - the meter reads low. Excluded by the sign of the error, but worth checking because a combined leak and some other error could masquerade as agreement.
Worn, reversed, damaged or wrongly sized orifice plate. Reads low; also this would have been present since installation, not appear eight months ago.
Flow computer configuration - square root applied to the percentage flow rather than the DP, or a wrong density or expansibility factor. Possible, but the error would have stepped in at a configuration change; check the configuration audit trail and the last download date.
Unequal impulse line elevations, lengths or insulation introduced at the annual shutdown (a replaced line, new lagging, a pot substituted for a different size). Very plausible given the timing.

Why the bench calibration and equalise check prove nothing about this fault. The bench calibration verifies the transmitter in isolation. The equalised-manifold check connects the two legs at the transmitter: any level difference between the pots simply drains through the equalising valve until the columns balance, and the transmitter then reads zero by construction. The 0.2 % result therefore proves the electronics and sensor capsule only - it cannot detect unequal condensate levels. Worse, if the loop is returned to service without re-establishing the condensate in both legs, the act of checking can itself create the offset.

Practical confirmation: physically verify both pot levels (sight glass, rodding or drain-and-weigh), check the drain valves for passing, compare the line temperatures and insulation, and cross-check the steam flow against the condensate and feedwater balance at three loads on the same day.

3. What permanent correction and maintenance regime would you put in place, and how would you respond to the proposal to bias the flow computer by 5 %?

Read the worked answer

Reject the bias, in writing. A fixed multiplier is arithmetically correct at one flow only; it would hide the fault, distort the allocation of the steam bill between the three lines, and remove the signal that would warn of a growing problem. Recommend instead that the flow computer configuration is locked and any bias requires a documented engineering change.

Correction:

Isolate and drain both impulse lines; replace any passing drain or vent valve.
Refill both legs to a verified equal level, confirm with pot sight glasses, and re-zero the transmitter at operating temperature with the process at a known condition.
Make the two legs identical - same elevation, same length, same pot size, insulated together and lagged to the same standard. Unequal lagging alone creates unequal condensation rates and re-creates the offset.
Verify the repair at three loads against the heat balance and, if available, a clamp-on ultrasonic check meter. Accept the repair only if the agreement is within about 2 % across the range.

Maintenance regime:

Add the impulse lines and pots to a quarterly routine with a written procedure and an as-found record; a floating level in one leg is a slow, silent failure.
Add a monthly automated comparison of the steam meter against the condensate and feedwater balance with an alarm at 3 % deviation, so the next occurrence is caught in weeks rather than eight months.
Keep the DP transmitter in the calibration programme but add a five-point check that includes the static pressure and temperature effects, and record the 'as-left' zero with the impulse lines proved equal.
For a cost-allocation measurement, consider an independent second measurement (vortex or clamp-on ultrasonic) so that no single transmitter carries the whole billing argument, and document the uncertainty in the allocation basis.