SONIC-NOZZLE SYSTEM PROVES TURBINE METERS AT ACTUAL FLOW RATES

Jim Beeson
Arkla Energy Resources
Shreveport, La.

Arkla Energy Resources, Shreveport, La., has developed and is currently using a mobile sonic-nozzle meter prover that permits proving at the actual flow rates.

The sonic-nozzle proving equipment is in use on 3-8 in. gas turbine meters at meter station sites. The prover also contains a gas chromatograph which is used in actual mass flow computations.

K-FACTOR ERROR

This system has several advantages over earlier methods of proving gas turbine meters. Arkla previously proved its larger turbine meters capable of 60,000 actual cu ft/hr (acfh) with a Roots transfer prover capable of only 10,000 acfh.

This meant that the provings were at or near increments on the proving curve where the K-factor (pulses acf) might be in error for the turbine meter's normal flow rate. With the sonic-nozzle prover, Arkla can now prove a turbine meter at the rate of actual flow.

It is established that a turbine meter proven at 0 psig and then put into pressure service will exhibit a K-factor shift in the positive direction. Without the meter's being returned to the manufacturer, this shift cannot be accounted for.

The sonic-nozzle prover calibrates turbine meters under actual operating pressure, or densities, in lieu of atmospheric pressure, and it calibrates under actual operating temperatures.

Further, the sonic-nozzle system uses natural gas rather than air as the proving medium to eliminate any conversion errors created by proving in air.

Turbine meters with removable cartridges that are normally sent to a proving facility can now be proven on-line. Proving on-line eliminates installation problems such as shifting mechanical components which can alter the proving curve.

Arkla designed its mobile proving system to return the proving gas into the line rather than exhausting it to the atmosphere as other nozzle-proving systems do. Returning the gas also makes the system much quieter to operate.

The total turbine-meter system, not just the turbine meter, is also proven.

System errors in turbine-meter readings, caused by pulsation for example, can be detected by the mobile prover. This capability is not possible with any other method of proving. The economic justification for developing this proving system was the improvement in total turbine-meter measurement accuracy. An improvement of 0.5% accuracy will pay for the prover in less than a year.

DEFINITION, EQUIPMENT

Sonic-nozzle flow is defined as follows: As nozzle downstream pressure decreases, maintaining all other factors, conditions in the nozzle reach a point at which further decreasing the downstream pressure will not further increase mass flow at the nozzle throat. At this point, the gas is moving at the speed of sound at the nozzle throat. Called "critical flow" or "sonic flow" or "choked flow," the mass rate of the gas at the throat is very accurately defined. It was earlier thought that a 50% permanent static pressure drop between inlet and outlet was needed to reach sonic velocity.

However, it is now known that a pressure drop of as little as 5%, and typically no greater than 10%, can produce sonic velocity.

The accompanying box provides meter-prover specifications.

The prover (Figs. 1-2) is in a Chevrolet box van mounted on a three-quarter-ton frame with a 6,100-lb chassis. Partitions separate the computer room, the valve room, and the generator/air room to divide hazardous from nonhazardous environments.

The computer control system consists of an IBM XT with a 20-megabyte (mb) hard-disk drive, a graphics printer, and a remote terminal unit (RTU).

The RTU, which is the interface between the computer and the valve, has control outputs, status inputs, and analog inputs.

Control outputs switch solenoid pilot valves (one solenoid per nozzle). Status inputs from two reed switches per actuator indicate the position of the actuator.

A status input from the air compressor indicates when the air pressure drops below 75 psi, the minimum pressure at which the Digicell will operate.

The analog inputs represent turbine-meter temperature and pressure (two inputs), valve inlet and outlet pressures (three inputs), and valve-inlet temperature (one input).

The prover also contains a Daniel chromatograph (Fig. 1).

Use of the gas chromatograph allows the incorporation of the actual real-time gas components (replacing an approximate analysis) into the gas equation, something never before accomplished. This unit performs a complete stream analysis every 6 min and sends the results to the IBM computer via an RS-232 serial link.

The computer uses the molecular percentage to calculate two of the variables used in the mass flow equation. The chromatograph system includes a strip chart recorder for printing a chromatogram or a trend on any component of particular interest.

Two sections comprise the chromatograph system. The gas chromatograph and line probe are located at the outlet piping of the Digicell in the hazardous area. The computer room, containing the chromatograph controller and the chart recorder, is considered the nonhazardous area.

Two generators make the prover totally self-sustaining. A 750-w unit powers small loads like the chromatograph, computer, and RTU. A 7.8-kw unit powers the air compressor and air conditioner.

The prover has a power receptacle that can be used instead of the generators if a station has electricity. Surge protection is installed in the main power-distribution panels.

COMPRESSORS, SOLENOIDS

A 1-hp air compressor is mounted next to the generators. It provides 100-psi air to the Digicell solenoids to operate the nozzles. It also supplies air to operate small hand tools used to repair turbine meters.

Two 14-ft lengths of 4-in. flexible pipe (Fig. 2) are used to move gas into and out of the prover. When not in use, the pipe is stored in the valve section of the prover.

One piece connects the meter to the prover, and the other piece connects the prover back to the meter. The pipe has an operating pressure of 575 psi, and Grayloc fittings on each end connect the prover to a station. (The Grayloc fitting is a high-pressure, quick-connect type.)

The Digicell (Fig. 3) weighs 1,800 lb and has 11 selective sonic nozzles sized in a binary progression.

The binary progression gives the Digicell 10-bit resolution because the largest two nozzles share the duty of the largest port through an "and/or" logic. This results in flow increment steps of 58.7 acf h.

The explosion-proof electrical enclosure at the top of the valve houses pilot solenoids. Electrical control signals from the computer activate the pilot solenoids which in turn operate the air-operated solenoids to put a nozzle in service. The air reservoir in the center area will cycle the Digicell at least twice if the air supply is lost.

The Digicell also has four holes tapped into its body, two in the inlet section and two in the outlet section. The tabs are used for inlet temperature and inlet and outlet pressure transmitters.

Since the inlet pressure is most critical, one inlet plenum tap has two pressure transmitters, with an accuracy of 0.10%. Their ranges are 0-50 psig and 0-200 psig. A drain is located at the base to blow down any liquids that might accumulate in the valve.

Once a nozzle is open, gas travels from the inlet plenum up through the nozzle, out four holes in the cage tube, into the outlet plenum, and out of the valve. Two reed switches per nozzle verify whether each sonic nozzle is open or closed.

The computer room contains two more pieces of equipment. One is a small oscilloscope used to observe the signals coming from the turbine-meter pulser. The other is a calibration verifier which simulates temperature and pressure signals to the computer and verifies the software is working properly.

Both of these devices are used primarily for troubleshooting the system.

STATION MODIFICATIONS

Arkla's current turbine-meter stations have no provisions to accept the mobile prover.

Selected high-volume turbine meters were modified on the downstream side with as much of the existing equipment as possible.

All piping modifications were made with 4-in. pipe, which has a negligible pressure drop on larger size stations, considering our normal operating pressure and flow.

The first items added to the station were two 4-in., full-port valves (Fig. 4) to route the flow to and from the prover. They were placed perpendicularly to the downstream piping and remain closed until the station is tested.

Next a full-port, double-block-and-bleed valve, the same diameter as the piping with a 4-in. minimum, was installed on the line between the two 4-in. inlet and outlet valves. It remains open until a station is tested.

The technicians use this type of valve to ensure all the gas that goes through the meter goes through the prover. Any gas not flowing through the prover that flowed through the meter nullifies the test.

To measure downstream temperature, we added a thermowell on the outlet side of the meter. A pressure tap was also made at the meter to measure turbine-meter case pressure.

These two signals are transmitted to the RTU which furnishes actual meter-operating conditions.

We installed Grayloc fittings at the 4-in. inlet and outlet valves, for quick connection. Grayloc fittings are clamped instead of flanged.

A by-pass around the turbine meter and the prover was added. The bypass also contains a full-port valve that bypasses some of the gas around the meter and prover. The by-pass is used to keep customers on when the Digicell starts cycling down flow through the meter.

A pulse output device or chopper disk was installed on the follower magnet. It is below the change gears and furnishes a varied number of pulses per revolution depending upon the meter size.

These modifications cost approximately $8,000/station.

FLOW EQUATIONS

The prover calculates a mass flow based on the following equation:

[SEE FORMULA (1)]

where:

M = Actual mass flow rate, lb

P = Plenum pressure, psia

A Area of the nozzle throat, sq in. Our nozzles range from 0.0620-0.9941 in. diameter, so that our nozzles range from an area of approximately 0.003 to 0.776 sq in.

C* = Critical flow factor, a function of the ratio of specific heats (isentropic flow coefficient). It is a parameter that relates the speed of sound of a gas with the flowing conditions of a gas.

Typical natural gas in Arkla's system has an approximate critical flow-factor value of 0.70 (unitless), and its value is pressure and temperature-dependent.

C* is calculated with chromatograph calculations of molecular percent of the gas diluents.

Cd = Discharge coefficient or actual mass flow rate divided by theoretical mass flow rate. In simple terms, it is the efficiency of the design of the nozzle.

Cd differs for each nozzle and is determined during calibration. Arkla's nozzles range from a discharge coefficient of 0.96 to 1.0 (unitless).

R = Gas constant = 48.03 the molecular weight of the gas stream. The gas chromatograph calculates the molecular weight of the gas stream used in this variable.

T = Plenum temperature, R. (F. + 460)

Volumetric flow rate is simply mass flow rate divided by density. Density, which is equal to P/(Z - R - T), can be eliminated from the formula to yield the following:

[SEE FORMULA (2)]

where:

V = Volumetric flow

Z = Plenum compressibility factor (R. C. Johnson,1 not American Gas Association) and has a value that approaches unity. The chromatograph calculates the molecular percent of the gas components used in this variable.

Rv Gas constant 2.398 the molecular weight of the gas. The chromatograph calculates the molecular weight of the gas stream used in this variable. R and Rv have been adjusted for engineering unit conversion.

PROVING FLOWCHART

Gas flows (Fig. 5) through the turbine meter then through the Digicell, generating four signals.

Two of the signals are meter-gas temperature and pressure. The other two are pulses from the turbine meter slot-sensor pulser and mass-flow rate from the Digicell. Note that the slot-sensor is positioned below the gears.

The computer knows the number of disc slots on the turbine-meter shaft so it can determine when the shaft has made a revolution.

The first set of gears, called intermediate gears, has an input-to-output ratio of 122.0555 to 1 for Rockwell meters and a varying ratio for American meters. They affect vertical shaft rotation significantly.

The second set of gears, called change gears, has an input-to-output ratio of between 1 and 2 to 1.

The index, or counter, is located atop the change gears and indicates vertical shaft rotation as 100 cu ft or 1,000 cu ft, depending on the size of the meter.

The index has a 100-cu ft drive on 3-in. through 6-in. meters. Therefore, each revolution of the shaft positioned above the change gears represents 100 cu ft. On 8-in. meters, the index has a 1,000 cu ft drive.

The final result of the slot sensor, intermediate gears, change gears, and index assembly is actual cubic feet.

The pressure-volume-temperature (PVT) chart is positioned atop the index and is used as a check for the index.

The computer inputs the meter temperature and pressure and applies it to the meter acfh value to produce standard cu ft/hr (scfh) at the meter.

Mass flow (in pounds) that the computer generates from the Digicell information, divided by the density at standard conditions, results in mass flow in scfh.

Mass flow in the system, which is the gas that flows through the turbine meter and then through the Digicell, remains the same at all points in the system. Volumetric flow in actual cu ft, however, varies according to composition, temperature, and most of all pressure of the gas.

Therefore, the volumetric flow at the valve will be different from the volumetric flow at the turbine meter. For a given mass flow, as the gas density increases the volumetric (acfh) flow decreases.

Volumetric flow at the valve is higher than volumetric flow at the turbine meter (because the inlet pressure to the valve is lower than the pressure at the turbine meter so the flow speeds up slightly) by the inverse of the ratios of the absolute pressures and an adjustment for supercompressibility.

That is why temperature and pressure are measured at both the valve and the turbine meter.

The computer simulates what the index would read if the gears were there. The difference in the turbine meter index scfh and the Digicell scfh is the turbine meter error for that flow. The difference is plotted as percent error vs. flow rate (Fig. 6).

This percent meter-error is plotted with actual flow rates for a specific K-factor or pulses per acfh. The pulses per acfh or K-factor can be shifted simply by entering the numbers for a different set of change gears into the computer.

As a result, a turbine meter can be proven as it is, and by application of different numbers to the software, it can determine which set of change gears is needed for the lowest error.

Using the software saves time and labor costs of proving the meter over physically changing the change gears and reproving the meter until the appropriate set is found. Up to ten different flow rates can be plotted on any particular graph (Fig. 7).

A positive meter error means that fewer acfh are flowing through the pipe than are being measured by the turbine meter with its present set of change gears (meter is "fast"); therefore, the customer is being charged too much.

SYSTEM SETUP, OPERATION

The prover is used on large industrial customers or town borders. A prime concern is not to deprive the customer during proving. To offset the possible 15% pressure reduction the prover might cause, the pressure regulation is increased by 15%.

If commercial power is unavailable for the prover, the two generators are started followed by the air compressor.

Next, technicians confirm that the chromatograph is ready to go on line; then they turn on the computer and associated equipment.

Next the meter pulser and temperature and pressure transmitters are installed at the turbine meter.

Technicians install the flexible pipe to and from the proving connections at the meter (Fig. 2). Then the inlet and outlet valves are opened and the double-block-and-bleed valve (Fig. 4) is closed, forcing flow through the Digicell which must be completely open.

Now the technician keys in various information about the particular station (station number, meter size, location, minimum downstream pressure allowed, number of desired proving points, etc.).

The technician determines the minimum allowed downstream pressure that the station must maintain without starving the customer. If the computer senses a lower pressure, it sounds an alarm.

The start sequence is given and the prover begins to ramp the Digicell to the requested flow rate.

The maximum flow that a turbine meter can be proven at is normally the flow of the meter when the technician arrives, making scheduling very important.

At some locations, Arkla can prove a turbine meter at higher flow rates than those at which the meter is currently flowing. One method is to "de-pack" or let the pressure drop downstream of the meter then run a test and let the line pressure build back up as the meter is being proven.

Another method at stations with two turbine meters in parallel, sharing the load, is partially or totally to shut down one turbine meter and put the load on the other for proving. Proving any point requires 1 min.

The computer will ramp the Digicell to the second test point. Here the technician may be required to open the station bypass enough to maintain the required downstream pressure. Meter error and percent flow are logged again.

This process continues until the computer completes the number of test points requested, normally 10. The whole process takes approximately 15 min and depends on problems that might occur.

The computer prints a complete proving report (Fig. 6) and a graphic turbine-meter curve report (Fig. 7).

Now the technician can review the proving curve for differences from preceding tests, which are also stored on disk. He can investigate any severe variations to see if something has malfunctioned with the turbine meter. After repairing the damage, he can prove the meter again redefining a new curve.

The technician also has the ability to see how another set of change gears would affect the curve without opening the meter. He enters a new set of change gears into the computer then recalculates the results.

Once he finds the set of gears with the lowest error, he can simply replace the gears on the meter with the size indicated by the computer.

In most cases, the turbine-meter error does not shift. This ends the test.

The technician completes the necessary paperwork and saves the results on a disk. The technician then returns the station to normal operations.

TEST RESULTS

Arkla has acquired large amounts of data with the prover. The following six figures are samples of these data.

Fig. 8 sets the standard for the remaining figures. It shows six turbine-meter test curves and a seventh which is the average of the six tests. Each test covers 10 different flow points. The average curve represents all 60 flow points covered in the six tests.

All six tests were performed at the AT&T plant in Shreveport on Nov. 2, 1988, with a meter pressure of 45 psig.

Although all remaining figures show only the average curve, each nonetheless represents six different 10-point tests.

Fig. 9 displays two curves which represent tests made at the AT&T plant in Shreveport, one performed on Nov. 2, the other completed on Nov. 7 (12 tests total). The meter pressure for both tests was 45 psig.

The curves are identical except for the point at 975 scfh flow rate. Because the curves are basically identical, the meter demonstrates repeatability.

Fig. 10 displays two curves which represent tests made at different meter pressures on Nov. 9, 1988. Again the curves were identical with the exception of the 2,400 scfh rate. There the 42 psig was slightly higher.

Fig. 10 also shows that the transfer prover stops at 10,000 scfh.

Fig. 11 displays five more curves representing tests at three different locations with flow rates considerably higher than can be tested with a transfer prover. The curves cover 30 tests total or 300 points.

The transfer prover stopped before the flow rate of the third test was reached. The number of points covered in the tests shows the sonic-nozzle prover's large range and the amount of additional information which now can be gathered with it.

The three sonic-nozzle curves, Greely Gas, Caldwell, Kan.; Morton Salt plant, Hutchinson, Kan.; and the High Plains gasohol plant, Colwich, Kan., show that meter error increases considerably past the 10,000-scfh mark, a fact that would not be shown in a test with a transfer prover.

Compare the sonic-nozzle curve (tested at 155 psig) with the transfer-prover curve (TP, tested at atmosphere) of tests at Morton Salt. At the normal flow of 118,000 scfh, the sonic-nozzle curve shows 1.71% error vs. the transfer-prover curve which would have to be interpreted as 1.02% error.

This means that if Arkla used the transfer-prover number there would be a 0.68%, or 802 scfh, error in measurement.

Now compare the sonic-nozzle curve (2.32%) with the transfer-prover curve (1.5% of High Plains' tests at a normal flow of 37,000 scfh). Here Arkla would have an error of 0.82%, or 37,000 scfh, or a yearly error of approximately 2.66 MMcf.

Fig. 12 combines the curves from Figs. 9 and 10 and adds three transfer-prover curves of tests made on three different dates (420 tests or 4,200 points). The transfer-prover curves are considerably higher than the sonic-nozzle curves until the 6,030-scfh flow rate point.

We know sonic-nozzle curves represent true meter errors. Therefore, if AT&T's meter operated between 975 scfh and 2,400 scfh and meter-error results were calculated based on the transfer prover, Arkla could have an error in volumes between 10 and 55%.

The lower curve of Fig. 13 shows the original proving with change gears 46T/63T installed. At 102 Mcf/hr, the meter error was -1.48%. This meant that the meter was registering 1.5 Mcf/hr less than it should. Over a year's time, Arkla would lose 13,244.2 Mcf.

With the sonic-nozzle prover, the computer can be asked to choose the best set of change gears for a certain flow rate. The computer selected a set (upper curve 43T/58T) to put the proving curve as close to zero (0.34%) as possible at the 102 Mcf/hr flow rate.

REFERENCE

R. C. Johnson, "Calculations of the Flow of Natural Gas Through Critical Flow Nozzles," Journal of Basic Engineering (ASME), September 1970, pp. 580-589.