Flow meters: choosing and diagnosing

Why does my magnetic flow meter show flow when the pump is off, or jump all over the place at low flow?

Flow with the pump off is almost always electrical noise read as velocity. A jumpy low-flow reading is that same noise, plus a meter sized on line size instead of on flow.

The electrode pair puts out a few millivolts at full scale. Against that, a stray current path or a floating reference shows up as an apparent velocity of a few feet per second. Confirm the sensor body, the process liquid, and the transmitter all share one ground reference. Confirm the signal cable shield is grounded at one end only — the transmitter end on most manufacturers, though follow the manual for that meter rather than a rule of thumb. The sources are usually VFDs, welding machines, cathodic protection, and impressed-current systems.

The character of the reading splits the causes. A steady offset on a full, static line points at grounding, or at a zero that was trimmed while flow was present. A wandering, noisy reading points at stray current, entrained gas, or electrode coating.

Jumpy at low flow has a legitimate cause too. Down near 1 ft/s the signal is small and the noise is not, so a meter sized on line size instead of flow will always be noisy at the bottom. Most manufacturers' sizing guidance puts the working band at roughly 3–15 ft/s. Dropping one line size with a reducer raises velocity by the square of the bore ratio — about 1.4 to 1.8 times across most schedule 40 steps, and about 2.3 times on a 6-in to 4-in step. That moves the working range out of the noise. More damping only hides it.

Before any of that, confirm the line really is static. A leaking block valve, a thermosiphon loop, or backflow through an idle pump is real flow. The meter is right.

Why does my magnetic flow meter show an empty pipe alarm when the pipe is full?

An empty-pipe alarm on a full line almost always means the transmitter has lost its electrical reference to the liquid. Coated electrodes, conductivity below the meter's limit, or missing grounding. Not an air pocket.

Detection drives a small current through the electrodes and watches the impedance. Anything that raises electrode impedance looks like air.

Rule out conductivity first. Rosemount specifies 5 µS/cm or greater for the 8705, and that figure is typical of the standard product across manufacturers. Deionized water, clean condensate, and hydrocarbon-contaminated water can fall below it and alarm on a completely full line. Low-conductivity sensor designs reach roughly 0.05 to 0.1 µS/cm, and capacitance-electrode sensors such as Yokogawa's ADMAG CA are published down to 0.01 µS/cm. That is a model you order, not a setting you change.

• Electrode coating — grease, oil film, scale, or biofilm. Pull the sensor and inspect, or run the electrode cleaning function if the transmitter has one.
• Grounding — with no ring or ground electrode on lined or plastic pipe, the liquid floats and the impedance check has nothing to reference.
• Conductivity of the actual fluid, measured with a meter, not assumed from the data sheet.
• The empty-pipe trigger threshold itself. It is a configurable number and is frequently left at a default that is too tight for a marginal fluid.

Then confirm the pipe really is full. A meter at a high point, on a downward leg, or upstream of an open discharge runs partly empty at low rates, and the alarm is correct. Vertical with flow upward is the right mounting. Horizontal with the electrodes at 3 and 9 o'clock is second best.

Do I really need grounding rings on a magnetic flow meter, or is a grounding strap enough?

A strap is enough only when the adjacent pipe is metal, unlined, and in direct contact with the liquid. Lined pipe, plastic pipe, or any wafer-style sensor needs grounding rings or a ground electrode.

The two do different jobs. The strap bonds the sensor body to the adjacent flanges. The ring bonds the liquid to that same reference. Rubber, PTFE, or cement lining, or PVC or HDPE pipe, means the flange bolts carry no liquid path. The strap then bonds metal to metal and leaves the process fluid floating. That is the condition behind a drifting zero, noisy readings, and false empty-pipe alarms.

Order rings in a material compatible with the process. 316 SS is standard, with Hastelloy, titanium, and tantalum available for aggressive service. A built-in ground electrode is cheaper and works well on most conductive liquids, but a ring presents far more surface area and is the better choice on low-conductivity or marginal fluids. Wafer sensors have no flanges of their own and therefore no metal-to-metal path to the adjacent pipe at all, so they take rings too.

One case is not a field decision at the flange. Where the plant has cathodic protection or an impressed-current system, bonding the liquid at the meter can route that current through your rings. That needs an engineering review of the grounding scheme.

What does high drive gain mean on a Micro Motion Coriolis flow meter?

The transmitter is working harder than normal to hold the tubes at their target vibration amplitude. The usual cause is gas entrained in the liquid.

There is no universal normal value, and Emerson says so directly: different sensor sizes have different normal drive gains, and some larger meters sit close to 50% in clean single-phase service. What matters is the change from that specific meter's steady value. Take the baseline when you know the meter is full of liquid or dry gas. Emerson treats a spike of more than 40% above that baseline as a reason to go look for vapor, and so should you.

Any value below 100% means the tubes are still at their precise target amplitude and the mass flow reading is usable. At 100% the drive has reached the maximum power the sensor can supply inside its intrinsic-safety limit. Past that the amplitude falls off target and the measurement degrades. The reading normally goes erratic on the way up, before it pins.

Log drive gain against tube frequency and indicated density. The three move in different directions for different faults.

• Entrained gas or flashing: drive gain rises, indicated density falls and swings.
• Coating inside the tubes: deposit adds mass, so tube frequency drops and indicated density rises, both steadily.
• Erosion: tube wall is lost, so frequency rises and indicated density falls, again steadily rather than erratically.
• Drive coil, magnet, or pickoff: drive gain high while frequency and density both sit at their normal values.

Rule out the process first. Pump cavitation, a throttled or partly closed suction valve, a leaking seal drawing air, a line that is not fully packed, a control valve flashing upstream. Adding backpressure downstream of the meter often collapses the gas and returns drive gain to baseline — that fixes the reading and confirms the cause in one move.

Where the transmitter supports Smart Meter Verification, run it. The routine measures tube stiffness against the factory baseline to confirm the flow calibration factor has not changed, which is what separates a structural change in the sensor — erosion, corrosion, coating — from a process problem. It takes under 90 seconds and does not interrupt flow.

How do I zero a Coriolis flow meter and how often should it be re-zeroed?

Zero it full of process fluid, at operating temperature and pressure, with no flow. Re-zero only when a zero verification shows the meter has actually drifted. Never on a calendar.

Micro Motion's own instruction is explicit about the order. Bring the line to normal operating temperature. Fill the sensor completely. Stop flow by closing the downstream block valve first, then the upstream block. Confirm zero flow and thermally stable conditions, then run the transmitter's zero routine.

Closing upstream first can pull the tubes toward vacuum or draw vapor out of solution. A block valve that passes even slightly will bias the zero. That is the single most common reason a field zero comes out worse than the factory zero.

Run zero verification before you run zero calibration. Zero stability is a laboratory-derived number published for each sensor size and model. If the measured zero offset falls inside it, leave the zero alone — the factory zero was taken under laboratory conditions and is usually better than anything a field zero will produce.

Re-zero after something real changes: re-installation, a different mounting or new piping stress, a large shift in operating temperature, a change in tube coating condition. Emerson's own recommendation on interval is to inspect the zero seasonally through the first year of operation, then extend the interval based on how that installation's zero has actually behaved.

Zeroing to make the meter agree with a tank gauge is not calibration. It relocates the error into the bottom of the range.

Why does my vortex flow meter read zero when there is steam flowing through the line, and how do I set the low flow cutoff?

The flow is below the meter's minimum measurable velocity and the low flow cutoff has clamped the output to zero. Vortex meters stop producing a countable signal well above zero flow. An oversized meter on steam does this at normal operating rates.

Shedding needs turbulent flow. Counting the shed vortices needs signal amplitude. Those are two separate limits. On the Rosemount 8800D, published volume accuracy is ±0.65% of rate on liquid above Re_D 20,000 and ±1.0% of rate on gas and steam above Re_D 15,000. Below the stated limit the meter does not stop. Accuracy degrades linearly to about ±2% down to Re 10,000, then runs ±2% to ±6% from Re 10,000 down to Re 5,000, which is the minimum measurable meter Reynolds number.

On steam, Reynolds number is rarely what binds. Signal strength goes as ρV², and Rosemount publishes the limit directly: a theoretical minimum measurable velocity of √(36/ρ) ft/s with ρ in lb/ft³ — that is, a minimum ρV² of 36 — with a practical floor of 6.5 ft/s on gas and steam and 0.7 ft/s on liquid.

Work an example. A 3-in schedule 40 line on 100 psig saturated steam runs at about 0.257 lb/ft³. Re_D crosses 15,000 at roughly 2 ft/s and 20,000 at under 3 ft/s. But the meter cannot produce a countable vortex until about 12 ft/s. The signal limit is four times the Reynolds limit here, and it is the one that puts a zero on the screen. Run the sizing program at actual pressure and temperature before blaming the meter.

Set the cutoff just above the meter's minimum measurable flow at actual conditions, in the same engineering units as the flow output. Too low and the transmitter counts pipe vibration, water hammer, and nearby valve chatter as flow, which inflates the totalizer while the line is shut in. Too high and it truncates real production at turndown.

Test it directly. Close the block valve, let the line settle, watch the totalizer for 10 minutes. If it creeps, the cutoff is too low, or the meter is picking up mechanical vibration and needs a different orientation or better pipe support. Check the shedder bar for fouling and erosion damage while you are there, and confirm the density configuration matches the header pressure you actually run.

Why is my clamp-on ultrasonic flow meter showing low signal strength or no signal?

In order of likelihood: dry or displaced couplant, wrong pipe data typed into the transmitter, an unprepared mounting surface, wrong spacing or mode, or a pipe that simply cannot pass ultrasound.

Work it in that order and do not skip ahead.

• Couplant. Grease dries out, washes away, or squeezes out on hot pipe. Remove both transducers, clean the faces and the pipe, apply fresh couplant, and re-clamp firmly.
• Pipe entries. Outside diameter, wall thickness, pipe material and its sound velocity, and liner material and thickness all feed the spacing calculation. A wall entered as schedule 40 on schedule 80 pipe puts the transducers in the wrong place by a measurable amount. Measure the wall with a thickness gauge at the mounting location.
• Surface prep. Paint, rust scale, and a weld crown under a transducer kill the signal. Take it to bare, smooth metal.
• Mode. V-mode (reflect) is the default. Switch to Z-mode with the transducers directly opposite on large, lined, or attenuating pipe — Z-mode is a single traverse, so the acoustic path is half as long.

Some pipes will not work at all. Cast iron and ductile iron scatter the beam off the graphite structure. Concrete lining defeats it too, and so does any liner that has delaminated and left an air gap behind it. A line that is not completely full is hopeless, and so is a fluid carrying entrained gas or heavy solids. If the signal is strong with the pump off and collapses when flow starts, the problem is in the fluid, not the mounting.

What causes a control valve to pass flow even when the positioner shows 0% closed?

The positioner reports stem position, not seat contact. 0% means the feedback linkage reached its calibrated zero, which is not the same as the plug being on the seat.

Four causes cover nearly all of it. The positioner zero was set to a stem position short of the seat. The bench set or spring range no longer matches the actuator, so it runs out of force before seating. The seat or plug is wire-drawn, eroded, or holding debris under it. Or the valve was never specified tight enough in the first place.

Seat leakage is a published specification, not a defect. Under ANSI/FCI 70-2, Class II allows 0.5% of rated capacity, Class III 0.1%, and Class IV — commonly sold as "tight shutoff" — allows 0.01%. Class V is a volumetric formula and Class VI is a bubble count.

Diagnose it at the valve, not at the DCS. Put the loop at 4 mA — closed on an air-to-open valve, reversed on air-to-close — then read actual actuator pressure against the bench set and check stem travel against the indicator with a scale. Stem at the mechanical stop with flow continuing means the trim is passing, and no amount of calibration will fix it. Stem short of the stop means re-zero the positioner so the closed signal drives the plug fully home, and check that no configured minimum-position or travel-limit parameter is holding it off the seat.

A throttling valve is not a block valve. If the process needs positive isolation, that belongs on a dedicated on/off valve in series. Trim replacement belongs with a valve shop or the OEM.

Why doesn't my PLC totalizer match the flow meter's own totalizer, and should square root extraction be done in the transmitter or in the DCS?

Extract the square root in one place. Totalize in one place. The mismatch is nearly always double extraction, a different low flow cutoff, or the host integrating on a scan interval the meter does not use.

Set the transmitter to square-root output and let the host block extract as well, and the displayed value becomes the square root of the true flow fraction. At a true 50% flow the host shows 70.7%. At 25% it shows 50%. At 10% it shows 31.6%. Zero and 100% agree exactly. That is why the error survives a two-point loop check and runs for years. If nobody extracts, it runs the other way — true 50% flow displays as 25%.

Common current practice is to leave the transmitter linear in DP and extract in the host. Square-root extraction has unbounded gain approaching zero, so a transmitter that extracts amplifies its own noise and zero drift at the bottom of the range and needs a low flow cutoff just to stay readable. A linear DP signal also makes double extraction obvious the moment you compare mA against indicated flow.

The trade is resolution. At 10% flow a linear DP signal sits at 4.16 mA where a square-root output would sit at 5.6 mA, so the host's analog input needs the counts to resolve it. A digital or fieldbus value removes the question entirely.

• Low flow cutoff. If the meter cuts off at 2% and the host integrates everything above 4 mA, the two totals diverge on every startup and shutdown.
• Integration method. A PLC that multiplies the present rate by its scan time drifts against a meter that counts pulses or integrates internally at a much faster rate.
• Damping. Transmitter damping delays the value the host sees, so ramps and batches total differently.
• Units and rollover. Gallons versus thousand-gallons, per-minute versus per-hour, and a 16-bit register wrapping at 65,535.

Where the total is the deliverable — billing, allocation, mass balance, custody transfer — take the pulse or digital output from the meter and let the meter own the total.

How do I measure a 4-20mA loop with a clamp-on mA meter without breaking the loop?

Zero the jaw closed and away from any conductor, then clamp a DC milliamp process clamp meter around one conductor of the loop — a single wire, not the pair.

Select mA. Close the jaws on nothing and zero the meter. Open them and clamp one conductor, alignment marks seated, polarity arrow pointing in the direction of conventional current. Clamp both conductors of the pair and you read zero, because the two currents cancel.

The Fluke 771/772/773 mA process clamp measures 0 to 20.99 mA at 0.01 mA resolution, specified 0.2% of reading plus 5 counts, and 21.0 to 100.0 mA at 1% plus 5 counts. Enough to tell 3.6 mA from 4.0 mA. Not enough to write a calibration record against.

Read the number against the NAMUR NE 43 zones so it means something. 3.8–20.5 mA is live measurement. ≤3.6 mA and ≥21.0 mA are transmitter fault. 0.0 mA is a broken loop or a dead power supply. The bands between — 3.6 to 3.8 mA and 20.5 to 21.0 mA — are differentiation zones a compliant transmitter does not sit in. A live 4 mA is 0%, not a dead loop.

A reading that wanders by more than a few hundredths of a mA wants a re-zero, and the clamp moved away from motor leads, transformers, and DC bus work. Fluke specifies the influence of the earth's own field alone as under 0.12 mA, so a real stray field near the jaw is worth more than that. For an accuracy verification that has to hold up on paper, break the loop and use a series-connected reference meter.

How many pipe diameters of straight run do I need before and after a flow meter, and can I use a flow conditioner to shorten it?

It depends on the meter type, the beta ratio, and what fitting sits upstream. A flow conditioner does shorten the requirement, but only within limits the standard and the manufacturer publish, not to any length you like.

Working numbers, upstream / downstream:

• Orifice plate. ISO 5167-2 Table 3, single 90° bend upstream, runs from 6D at β ≤ 0.20 to 44D at β = 0.75 in Column A, for zero added uncertainty. Column B — accept +0.5% added uncertainty on the discharge coefficient — runs 3D to 20D for the same fitting. Downstream is 4D to 8D in Column A across the beta range. The worst case in the table is two 90° bends in perpendicular planes, which reaches 75D at high beta, with a footnote requiring 95D above Re_D 2 × 10⁶ when the two bends are closely spaced. Look up your beta and your actual fitting; do not use a rule of thumb.
• Vortex. Rosemount publishes 35D upstream and 5D downstream for the 8800 with no K-factor correction, and calls 10D upstream the minimum recommended, where the K-factor may shift by up to 0.5%. Use that meter's own table rather than a generic figure.
• Magnetic. 5D up, 2D down, measured from the electrode plane. The most forgiving common meter.
• Ultrasonic, clamp-on or inline. 10–20D up depending on the disturbance, 5D down, and substantially more downstream of out-of-plane bends, control valves, or a pump. Single-path clamp-on is the most profile-sensitive of the group.
• Turbine. API MPMS Chapter 5 gives 10–20D upstream depending on meter size and 5D downstream, reducible with a straightening vane section.
• Coriolis. No straight run requirement — it measures mass directly and does not care about velocity profile. It does care about support and piping stress, which move the zero.

ISO 5167-2 spells out the limits on conditioners. A 19-tube bundle flow straightener can be used downstream of any fitting for β ≤ 0.67, but the fitting must still be at least 30D from the plate, with the straightener's downstream end at 13D ± 0.25D. To shorten the run below that, the standard's Table 4 route requires the nearest fitting to be at least 18D away. Below 18D the standard gives no location and will not predict the uncertainty.

The Zanker flow conditioner plate is the other unpatented option in the standard: β ≤ 0.67, nearest fitting at least 17D from the plate, and the plate's downstream face between 7.5D and L_f − 8.5D from the orifice. Annex B covers two patented conditioners that passed the same compliance test — the Gallagher conditioner, also at L_f ≥ 17D, and NOVA's K-Lab perforated plate. A four-hole conditioning orifice plate such as the Rosemount 1595 goes furthest, rated for 2D upstream and 2D downstream. It carries the manufacturer's uncertainty rather than the standard's, and it adds permanent pressure loss.

One note on editions. ISO 5167-2:2022 is the current second edition and replaces the 2003 edition. Its stated changes are a revised maximum orifice edge thickness for β < 0.2, a correction to the required spacing between two 45° bends, a clearer specification for the tee, and the addition of flow calibration. None of that touches the Table 3 numbers above. Work from the edition your measurement standard actually cites.

Which way does the sharp edge of an orifice plate face - upstream or downstream?

The sharp square edge faces upstream into the flow, and the beveled side faces downstream.

The discharge coefficient in ISO 5167-2 and AGA Report No. 3 comes from a sharp square edge forming the vena contracta. Turn the plate around and the flow meets the bevel first, the jet forms differently, and the meter reads low by a large, steady amount. NEL's test and CFD work on reversed plates puts the actual discharge coefficient about 22% ± 3% above the standard value — a correction factor of 1.22 on any flow derived the normal way, so the indicated flow runs roughly 15% to 20% below actual. Field reports on natural gas meters land in the same place, commonly 15% to 17% low. ISO 5167 itself does not quantify the error at all; the standard assumes correct orientation.

Nothing about the loop looks broken. The transmitter checks good, the DP is stable, and the number is wrong.

• Read the tab. It is normally stamped on its upstream face with the bore, the plate material, and the word INLET or UPSTREAM, or marked with a flow arrow. Read the words — do not assume the stamped face is the upstream face on an unmarked plate.
• Feel the bore. The upstream side is a square corner you can catch with a fingernail; the downstream side is a chamfer.
• Check the vent or drain hole separately. It is drilled at the pipe wall — vent at top for liquid, drain at bottom for gas or steam — and its position is a separate check from plate orientation.

Per ISO 5167-2, the plate is beveled on the downstream side only when its total thickness E exceeds the bore edge thickness e, at a bevel angle of 45° ± 15°. The limits are 0.005D ≤ e ≤ 0.02D and e ≤ E ≤ 0.05D. The upstream edge counts as sharp when its radius is no greater than 0.0004d, and square when the bore-to-face angle is 90° ± 0.3°. A bidirectional plate is not beveled at all, and both faces must meet the upstream-face specification. Those figures are the 2003 edition's. The 2022 edition revised the maximum edge thickness for β < 0.2, so check the edition you are working to if you are down at low beta.

How do I size an orifice plate bore for a given max flow and 100 inH2O DP range?

Solve the ISO 5167 flow equation for bore at maximum flow and 100 inH2O, then confirm the resulting beta ratio lands inside 0.10 to 0.75. Aim for 0.40 to 0.65.

For liquid in US units the working form is Q = 5.667 · C · E · d² · √(h_w / G). Q is gpm, d is bore in inches, h_w is DP in inH₂O, G is specific gravity, C is the discharge coefficient — about 0.605 for flange taps at moderate beta and high Reynolds number — and E = 1/√(1−β⁴) is the velocity of approach factor. β depends on d, so it is iterative. Guess β, solve for d, recompute β, repeat. Two or three passes converge.

Worked example — 250 gpm of water in 4-in schedule 40 pipe (D = 4.026 in) with a 100 inH₂O range. Iterating gives a bore of about 2.58 in at β ≈ 0.64. Check the answer before ordering the plate:

• Beta in range. ISO 5167-2 covers 0.10 ≤ β ≤ 0.75, with bore d ≥ 12.5 mm. Coefficient uncertainty is 0.5% for 0.20 ≤ β ≤ 0.60 and rises to (1.667β − 0.5)% above that — roughly 0.57% at β = 0.64. Below β = 0.20 it is (0.7 − β)%.
• Permanent pressure loss. ISO 5167-2 gives it as approximately (1 − β^1.9) of the DP, so about 57% of 100 inH₂O here — roughly 2.1 psi the pump has to give up.
• Reynolds number at minimum flow. Flange taps require 50 mm ≤ D ≤ 1000 mm, Re_D ≥ 5,000, and Re_D ≥ 170β²D with D in mm. Below D = 71.12 mm (2.8 in) an extra uncertainty term applies.
• Turndown. A single 100 inH₂O range is useful to roughly 30% of maximum flow, where DP has already fallen to 9 inH₂O and transmitter error starts to dominate.

The 0.605 above is a hand-calculation value. A production sizing uses the Reader-Harris/Gallagher equation for C at the actual Reynolds number. For gas, steam, or anything two-phase, add the expansion factor and run a proper ISO 5167 or AGA 3 calculation. The liquid form above will get you a wrong bore.

How do I calculate the K-factor for a turbine flow meter in pulses per gallon?

K-factor is total pulses divided by total gallons. In service the rate form is K = 60 × f(Hz) ÷ Q(gpm).

1,250 Hz at 100 gpm gives K = 60 × 1250 ÷ 100 = 750 pulses per gallon. To determine it in the field, divert into a calibrated vessel or run against a prover, count raw pulses over the run with a totalizing counter, and divide by the delivered volume. One run at one flow rate gives you one point. That is not a K-factor.

A turbine's K-factor is not a single number. It sags at the low end where bearing drag and viscosity dominate, sits flat across the linear range, and rolls off approaching maximum. That is why the calibration certificate lists K at several flow rates. Use the average K across the linear range, or load the multi-point curve if the flow computer supports it.

K shifts with viscosity as well — a meter certified on water will not hold that K on 30 cSt oil — and with bearing wear. A turbine that has drifted a couple of percent against a prover has a mechanical problem. The fix is bearings and a re-certification, not a new K entered in the DCS.

How do I calculate the Cv I need to size a control valve for a given flow and pressure drop?

For non-flashing, non-viscous liquid in a line with no reducers, Cv = Q × √(G / ΔP), with Q in gpm, ΔP the pressure drop across the valve in psi, and G the specific gravity.

250 gpm of water at a 10 psi drop needs Cv = 250 × √(1/10) = 79. Then size the valve so maximum required flow lands at 60–80% of travel and minimum required flow stays above 10–20%. A valve where max flow sits at 95% open has no margin left for fouling or a changed process. One that operates below 10% will cycle, cut the seat, and never control well. Compare your calculated Cv against the manufacturer's Cv-versus-travel table, not the rated Cv at 100% open.

• Reducers. A valve smaller than the line needs the piping geometry factor F_P, which divides into the available Cv. Skipping it undersizes the valve.
• Choked flow. Liquid chokes when ΔP exceeds F_L²(P₁ − F_FP_v), where F_L is the valve's liquid pressure recovery factor and F_F ≈ 0.96 − 0.28√(P_v/P_c). Past that point additional ΔP buys no additional flow, and the valve is cavitating or flashing.
• Gas and steam. Use the IEC 60534-2-1 compressible equations, with the expansion factor Y and the pressure drop ratio factor x_T. The US adoption is ANSI/ISA-75.01.01-2012, and it is a modified adoption — the standard is titled "(60534-2-1 MOD)" — so confirm which document your specification cites. The liquid formula undersizes badly at high pressure drop ratios.

Size on the ΔP the valve actually gets at maximum flow, not the pump's shutoff head. System friction rises with the square of flow, so the valve's share of the total pressure shrinks exactly when maximum capacity is needed. Flashing, cavitation, or high noise is a trim selection problem. That belongs with a valve engineer or the manufacturer's sizing group.

Coriolis vs magnetic flow meter - which one should I use, and what's the best flow meter for abrasive slurry lines?

Coriolis when you need mass or density, or the fluid is not conductive. A magnetic meter when the fluid is conductive, the line is large, or the service is abrasive. For abrasive slurry the answer is nearly always a mag meter with a polyurethane or ceramic liner.

The decision usually makes itself on one of four points:

• Conductivity. Hydrocarbons, solvents, deionized water, and gases rule out a mag meter entirely. The standard product needs at least 5 µS/cm.
• Mass or density needed. Blending, custody transfer, and reaction charging favor Coriolis. Micro Motion's ELITE line is specified at ±0.10% of rate on liquid mass, with ±0.05% available on premium sensors, plus a density measurement.
• Line size. Past about 6 in, Coriolis cost, weight, and pressure drop climb steeply. Mag meters stay reasonable to very large sizes and add essentially no pressure drop.
• Abrasion. Coriolis tubes erode and shift calibration. A full-bore mag meter has no obstruction in the flow path — only the liner and the electrode faces see wear.

For slurry, specify around the wear and not the accuracy number. Polyurethane gives the best abrasion resistance for sand, ore, and coal slurries, and it is temperature-limited: Rosemount publishes 140 °F (60 °C) for standard polyurethane and 158 °F (70 °C) for the extreme-service grade. Both are out on hot slurry. Alumina ceramic is harder still but will not tolerate thermal shock or impact from large solids.

Size for roughly 5–10 ft/s. Fast enough to keep solids suspended, slow enough that liner life is measured in years, and well clear of the low-velocity region where mag meters get noisy. Specify flush or bullet-nose replaceable electrodes.

Two exceptions worth checking before you order. Magnetite and other magnetic solids can distort the field. Very high solids loading generates its own electrode noise. Both are questions for the manufacturer's application group, and the answer may come back as a Coriolis meter or a non-contact technique instead.

Clamp-on vs inline ultrasonic flow meter - how much accuracy do you lose with clamp-on?

Expect ±1% to ±2% of reading from a clamp-on installed on a real pipe, against ±0.15% to ±0.5% from a multi-path wetted inline spool. Nearly all of that gap is uncertainty about what the pipe actually is.

An inline spool is calibrated as an assembly. Bore, path length, and transducer angle are all fixed and known at the factory. Multi-path inline meters are published at ±0.15% to ±0.5% of reading and single-path inline at ±0.5% to ±1%.

A clamp-on computes the acoustic path from numbers a technician types in: outside diameter, wall thickness, wall material and its sound velocity, liner material and thickness. Every error there scales the reading directly. Manufacturers do publish ±0.5% or better for clamp-on under controlled conditions, and that is not marketing — it is what the technology does when the pipe data is exact. The spread between that and what you get in a pipe rack is installation-driven, not inherent.

Wall thickness is the worst offender. It is stamped as nominal, and ASTM A53 and A106 both allow the wall to run 12.5% under nominal at any point before it leaves the mill, and that is on top of whatever corrosion has taken since. Measure it with an ultrasonic thickness gauge at the actual mounting location and enter what you measured. That one step usually buys more accuracy than anything else available to you.

The rest of the error is velocity profile. A single-path clamp-on samples one chord and assumes a fully developed profile, so any upstream disturbance lands directly in the reading. Dual-path or cross-path mounting recovers a good part of it.

Clamp-on earns its place by needing no shutdown, no cut-in, and no flange. For pump curve checks, system balancing, leak surveys, and verifying a suspect installed meter, ±2% today beats ±0.3% after the next turnaround. For billing, allocation, or anything referenced to an API or AGA measurement standard, a clamp-on is the wrong instrument and Bridges Industrial will say so.

Thermal mass flow meter vs Coriolis for compressed air - which is more accurate, and why does thermal read wrong when the gas composition changes?

Coriolis is more accurate: roughly ±0.25% to ±0.35% of rate on gas, against about ±1% of reading plus ±0.2% to ±0.5% of full scale for thermal. Thermal is still usually the correct choice on plant compressed air. It reads wrong on a composition change because it infers mass from heat transfer, and heat transfer is a property of the specific gas.

A thermal meter measures how much heat a flowing gas carries away from a heated element, then converts that to mass flow using the gas's thermal conductivity and specific heat. Change the gas and those properties change. A meter calibrated on air and used on something else reads wrong by whatever the property ratio is.

The accuracy gap is real but modest in absolute terms. Micro Motion publishes ±0.25% and ±0.35% of rate gas options on ELITE, while insertion thermal meters are typically ±1% of reading ±0.2% of full scale on air and ±1.5% of reading ±0.5% of full scale on other gases. Air and nitrogen are close enough that the shift is small. Air versus carbon dioxide, methane, or hydrogen is not close at all.

On compressed air the version that bites is moisture. A meter calibrated on dry air sitting downstream of a failed dryer reads off, and liquid water reaching a hot element produces a large false flow spike.

Thermal still wins on plant air for practical reasons. It goes in as an insertion probe through a hot tap without cutting the header, adds essentially no pressure drop, and reads directly in scfm without a separate pressure and temperature measurement. A 4-in air header does not justify what a 4-in Coriolis costs, weighs, or takes out of discharge pressure.

The catch with an insertion probe is that it samples one point. It needs the same straight run a full-bore meter would want, and it needs the insertion depth and orientation the manufacturer specifies — typically the probe tip about an inch past the pipe centreline, which puts the sensing element on the centreline. Get either wrong and the profile assumption behind the reading fails.

Reach for Coriolis when the accuracy is the requirement: custody transfer, a stream whose composition genuinely varies, an application where you need density as well as mass. And if the gas composition swings and neither meter is calibrated for it, the meter needs re-ranging by the manufacturer for the actual gas. Field trimming will not do it.

Venturi tube vs orifice plate - which one has less permanent pressure loss?

The venturi, by a wide margin. ISO 5167-4 puts a classical venturi's permanent loss at 5% to 20% of its differential, where an orifice plate loses 42% to 95% depending on beta ratio.

The difference is the diverging cone. A venturi's gradual expansion converts velocity head back into static pressure. An orifice plate has no recovery section at all, so the jet expands abruptly and dissipates the energy as turbulence and heat.

ISO 5167-2 gives the orifice approximation as Δϖ/Δp ≈ 1 − β^1.9, which works out to about 42% at β = 0.75, 62% at β = 0.60, 82% at β = 0.40, and 95% at β = 0.20. ISO 5167-4 states the venturi's relative loss can generally be taken as 5% to 20%, and gives the direction of every influence. It falls as beta rises and as Reynolds number rises. It grows with the divergent angle and with roughness. The standard permits a divergent included angle of 7° to 15° and recommends 7° to 8°. A steep cone is where you hand the recovery back.

Cost and installation flip the comparison. A venturi costs several times an orifice assembly, is heavy, is a fabricated spool that must be cut into the line, and cannot be re-ranged by swapping a disc during a turnaround.

Choose the venturi when the permanent loss is a real operating cost — a large cooling water line, a gravity or low-head system, a compressor suction, a pump with no head to spare — or when solids would build up against a flat plate. Choose the orifice when head is available and the range may change. A flow nozzle sits between the two. On a retrofit with short straight run, price a cone meter or an averaging pitot before you assume those are the only two options.

What type of flow meter is best for saturated steam?

A vortex meter with integral pressure and temperature compensation is the default for saturated steam. An orifice plate with a separate pressure transmitter is the low-cost alternative, and Coriolis is the answer when you need better than about 1%.

Saturated steam sits on the saturation line, so one measured pressure gives density straight from the steam tables. You do not need both P and T for density. You do need the pressure measured rather than assumed, and that single point matters more than the meter choice. At 100 psig saturated steam is 0.257 lb/ft³ (337.9 °F); at 110 psig it is 0.278 lb/ft³ (344.2 °F). Run a fixed density while the header floats 10 psi and roughly 8% goes straight into the mass total, every hour, forever.

Watch for superheat. Downstream of a pressure reducing valve the steam is no longer on the saturation line, and pressure alone will give you the wrong density. There you need both P and T.

• Vortex: no moving parts, and high-temperature versions of the Rosemount 8800 are rated to 800 °F (427 °C) — check the model you actually have. With full pressure and temperature compensation it reads mass directly at about ±1.2% of rate at 150 psia, degrading to ±2.5% at 2,000 psia. Its weakness is the bottom of the range: the minimum measurable velocity is set by ρV², so it climbs as header pressure falls — roughly 12 ft/s at 100 psig and about 22 ft/s at 15 psig. Size on actual flow, not on line size, or the meter will sit below its own floor at normal rates and read zero.
• Orifice with DP and pressure: inexpensive, well understood, repairable, and every technician in the plant can troubleshoot it. A single DP range gives about 3:1 turndown in flow, because a 10:1 span in DP is only √10 in flow; a stacked or multivariable transmitter does better.
• Coriolis: accurate and independent of any density assumption, but tube size, pressure drop, and cost keep it to small utility and metering lines.

Whichever you pick, most steam metering trouble is impulse-line trouble. On a horizontal run, take the taps from the side, at or just above the horizontal centerline. Top taps belong on dry gas and will not hold a condensate seal. Slope the lines down to a transmitter mounted below the taps. Use condensate pots filled to matched levels. Heat-trace only where freezing is a real risk, and control the trace so the seal legs cannot flash.

Wet steam degrades every steam meter — the water fraction bears no relation to the vapor density you configured. If the header carries condensate, fix the traps and the separator before you believe any number the meter produces.

When this page runs out

This page covers the failure modes that repeat. It will not cover the one in your line right now if the process is doing something unusual, and it is no substitute for the meter's own manual, ISO 5167, AGA Report No. 3, or the manufacturer's sizing program run at your actual conditions. Vendor figures quoted here are named to the meter that publishes them, because the next model over will not match.

Some of this is not field work. An accredited calibration certificate, a repair under an OEM warranty, a custody transfer measurement referenced to an API or AGA standard, a slurry or two-phase application that needs a laboratory behind it — those belong with an accredited calibration laboratory, the manufacturer, or a measurement specialist, and Bridges Industrial will say so. Where the problem is a meter that was never matched to the line, or one installed, ranged, or configured in a way the application does not tolerate, that is field work and we do it.