Presented at the University of Warwick, England, 1981



There is discussion of the objectives for the design of a 6000 m3/hr custody transfer metering system for refrigerated N.G.L. (propane and butane), used as part of a ship loading facility, and the constraints which shaped the final design. An outline of the advantages and disadvantages of different flow metering and meter proving possibilities leads to a description of the chosen meters and metering philosophy, a philosophy based on the use of full line sized, 24 inch, ultrasonic and vortex flow meters in each product loading line, proved by comparison with storage tank measurements. A description of the necessarily extensive instrument system in, and around, the tanks is supported by the results of a comprehensive accuracy analysis used to guide the metering system design and estimate the uncertainty in ship loading measurements.




Today there are many natural gas liquid (N.G.L.) plants producing refrigerated, liquid propane and butane for transportation to consuming areas. At present the ship loading terminals rely on the use of calibrated ships to determine the quantities dispatched, but there is a growing desire to measure cargoes by means of shore based facilities. To do this with an accuracy comparable to that usual for other hydrocarbon products presents some difficulties because of the non-lubricating, refrigerated nature of N.G.L's, the usual practice of continually cooling the loading lines up to a point near to the point of loading and the need to regularly prove the flow meters in situ. The problems are aggravated when the facility is large but, as has been recently proved, they are not made insurmountable. A series of innovative steps have led to a novel solution for the design of a custody transfer, refrigerated metering system of unprecedented size.


The subject metering system is an addition to a fairly typical N.G.L. storage and loading terminal designed for Ruwais in Abu Dhabi. The terminal was already under construction when it was decided to add a custody transfer metering system and it became an objective that the addition of this new facility should not delay the terminal's start-up --- even if the metering system was not fully completed.

The basic features of the loading terminal, as originally conceived, are illustrated in Fig. 1. Liquefied propane, at -43 deg.C, and liquefied butane, at -5 deg.C were each to be pumped down a 2 Km pipeline from a refrigeration plant to storage tanks near the sea shore. There the product was to accumulate in one of two 100,000 m3 storage tanks prior to being loaded on board ships moored at the end of a 2 Km long jetty. The jetty ran directly out to sea from a point near the storage area, and carried, for each product, a 24 inch loading line, an 8 inch recirculation line, a 30 inch vapour return line, a surge drum and loading booms. The recirculation lines, branching off the loading lines near to their seaward end, were to allow the loading lines to be kept cooled by means of a continuous circulation of liquid product from storage tanks, down the jetty and back to the storage tanks.

The terminal was to be capable of loading liquid propane and butane simultaneously at flow rates of up to 3000 m3 per hour each. However, general operational requirements and the planned acceptance of ships with much less than this maximum loading capacity meant that, for short periods of time, loading rates were expected to sometimes drop as low as 200 m3 per hour.

In designing the new metering system certain constraints had to be taken into account.

(i) In order to reduce the possibility and latitude for argument that there could be any difference between what would be metered and what would be loaded on board ship, the flow meters on which the sales were to be based needed to be downstream of the recirculation line and as near the loading booms as possible.

(ii) Any modification to pipe work on the jetty needed to be physically small and introduce minimal extra pressure drop into the loading line pressure profiles.

(iii) The flow metering system had not to interfere with the storage and loading facility's originally planned capacity, flexibility or safety.


Different metering possibilities were considered. Some were immediately dismissed. Differential pressure flow meters with their limited range and low accuracy, were judged unsuitable for the application. Positive displacement meters were considered too big, heavy and of undesirable design for use on volatile, refrigerated, non-lubricating N.G.L's. However, certain devices seemed to have interesting possibilities and so are discussed here in more detail.

3.1 Turbine meters

The use of turbine meters, with their wide acceptance as custody transfer measuring devices, was the obvious first choice, but on further consideration proved not to be so attractive. On refrigerated propane and butane they were known to hold their typical repeatability of better than +/- 0.1 % for a linearity of +/- 0.25 % over a range of 8 : 1. However, to cover the loading facilities flow range requirement of 15 : 1 (3000 to 200 m3/hr), and provide some measure of redundancy, the minimum requirement was two series pairs of 16 inch meters, with a valved cross-link between the mid. point of the two series paths. This was an unattractive option. It needed modification to the jetty pipe work and the use of flow straightners, which would add further pressure drop to that produced by the extra bends, valves and flow meters also required. Further more, it was uncertain whether quotations would be obtained for such large flow meters to be used under the planned operating conditions and, if so, whether they would meet the desired metering performance and withstand pressure waves induced by an emergency shut down of the loading line. Such large turbine meters had not previously been made for use on refrigerated N.G.L.

A further problem was how to prove the meters. Because of the low temperature of the liquefied propane, and the fact that proving would need to be done on a fairly regular basis, the only possibility was to use a large piston prover. This would involve moving into an area of untried technology, since only medium sized provers of this type had previously been built. Again there would be the problem of obtaining quotations and, if obtained, the uncertainty of the prover's satisfactory operation after installation. Accommodating the prover on the jetty would have also presented a major problem.

A possible alternative use of turbine meters was the more conventional approach of using a battery of small meters in parallel to at least enable manufacturers to work in areas of known technology as regards meter and prover size. However, substantial modification would have been needed to the jetty and pipe work to be able to accommodate so many parallel meter runs, let alone a prover. The increased pressure drop down the loading line resulting from such modification, would reduce maximum loading rate, and there would still be the problem of likely damage to turbine meters in the event of an emergency shut-down.

3.2 Vortex Shedding Meters

Vortex shedding meters offered an interesting alternative. Full line size (24 inch) instruments, with a claimed +/- 0.5 % repeatability over a flow range of greater than 15 : 1 were available, as were versions fitted with easily removable vortex sensors to avoid the need to install the meter between closely mounted block valves to facilitate sensor removal for maintenance.

Modification to the loading line to effect the meter's installation would only require the removal of an equivalent length of straight pipe and the fixing of flanges. Since this would necessitate neither modification to the line direction nor the close positioning of block valves, the existing long straight lengths of pipe along the jetty could be used to considerable advantage. Flow straightners would not be needed and the total extra pressure drop caused by the meter's installation would be only that caused by the meter itself, a fairly modest two velocity heads. Also, it was judged, the simple, rugged structure of the instrument would render it unlikely to suffer from a pressure surge caused by an emergency shut-down.

On the debit side, 24 inch vortex shedding meters had not previously been used on refrigerated N.G.L., but smaller versions had, and there was no reason to believe that the known good performance of the smaller meters would not be reflected by the hitherto untried larger ones.

Much more serious was the problem of how to prove large vortex meters in the field. The instrument's manufacturer could provide 'meter factors' based on water calibration, and these could be expected to hold good over long time intervals, but experiments with small vortex meters operating on a variety of liquids suggest that the calibrations on propane or butane would be slightly different to that applicable to water, (Ref. 1). If a piston prover was to be used to any worthwhile effect, the 24 inch vortex meter's very low frequency and irregularity of vortex shedding would require a prover considerably larger than that necessary to prove an equivalent sized turbine meter. There was no hope of being able to obtain, let alone install, such a device.

3.3 Diagonal-beam ultrasonic flow meters

Another interesting possibility was the use of ultrasonic flow meters of the diagonal-beam type, i.e. the type which determines flow rate from the time of flight of acoustic pulses projected diagonally across the flow stream. The best of these types compute flow rate from times obtained along several diagonal chords, and are claimed to be capable of a repeatability of better than +/- 0.25 % over a range of greater than 15 : 1. Added to this, the large pipe sizes and low viscosities of both water and the N.G.L's would ensure very large Reynolds numbers, even at low flow rates, so that a water calibration of the instrument could be expected to hold fairly good for the N.G.L. products, a point which would be of considerable advantage when first using the metering system.

The availability of 24 inch flow meters of this type, with transducers which could be removed from the instrument while under line conditions, offered the easy installational and maintenance possibilities described earlier for vortex meters. The ultrasonic flow meter's total lack of obstruction to the flow meant that they wouldn't add additional pressure drop in the loading line presser profile, and there should be no problems resulting from an emergency shut-down. Again, as with vortex meters, the large size instruments had not previously been used on refrigerated liquids, but they were confidently expected to behave like smaller meters of similar design --- which had proved to work well on refrigerated products. As with vortex meters, there was the major problem of how to prove the ultrasonic flow meters in situ.


It was finally decided to install, in each loading line, in series, both a line size diagonal beam ultrasonic flow meter and a line size vortex meter. This arrangement would minimize modification to the pipe work, could be easily implemented, and would add negligible extra pressure drop to the loading line pressure profile.

The reason for using two flow meters in series was partly to provide redundancy in case of a malfunction by one meter. The reason for using two dissimilar flow meters was more subtle. It was argued that the two instruments, being very different, would have different failure modes and calibration drift characteristics so that the drift of either one would quickly reveal itself to indicate the need for remedial action.

With both types of flow meter expected to exhibit good long term calibrational stability, and there being a high probability of getting an early indication of meter drift, there was no longer a need to provide regular, quickly available proving facilities. Accordingly it was decided that a mere upgrading and supplementation of tankage measuring facilities would surfice to render the tanks themselves suitable for calibrating the flow meters.

It total, therefore, it was decided to meter with ultrasonic flow meters, check with a vortex meters, and calibrate these custody transfer meters from a mass balance round the tanks using an improved instrument system in and around the tanks.


The location of the loading line flow meters had to satisfy several partially conflicting requirements.

(i) They had to be as near as possible to the ship's loading arms, since that would be the point of custody transfer, and there had to be no valves or branches between meter and loading arm to allow metered liquid to escape being loaded on board ship.

(ii) The meters had to be given good upstream flow conditions to perform well which, to avoid using flow straightners, meant locating the meters downstream of quite long lengths of straight pipe.

(iii) The meters had to be accessible for service and removal by jetty mounted lifting gear.

(iv) The meters had to be continuously at the temperature of the refrigerated product.

(v) The meters had to be installed in a length of line which, if necessary, could be safely cleaned of product and purged to allow their removal.

In order to satisfy all these requirements it was decided that the meters should be located near the end of the jetty, downstream of a long length of straight pipe, but upstream of a 24 inch to 18 inch cone, a series of bends, the loading rate control valve and, of course, the loading booms, Fig. 2. To keep the meters filled with refrigerated liquid a minor modification was made to the pipe work to allow their inclusion in the recirculation cooling loop when ship loading was not in progress, and yet outside of it when being used for metering. The two states were securely defined by interlocked double block and bleed valves, Fig. 2.

Flow meter removal facilities were provided by locating a full bore gate valve one hundred pipe diameters upstream of the first flow meter in each pair, together with line drainage and purge connections to allow the line as far back as the gate valve to be cleared. It was decided to install the ultrasonic and vortex meters fairly close to each other with the ultrasonic meter, which creates no flow disturbance, upstream of the vortex meter.


Because N.G.L. product is metered by volume, but sold by mass, the associated measurement of density was of crucial importance. There are two main ways by which the density of an N.G.L. product can be determined; the PVT method, by which the temperature and pressure of the liquid is measured and the density derived from tabulated values (or a corresponding algorithm) of the liquid's properties or, by direct measurement using a densitometer.

When applied to a pure liquid, whose PVT properties are accurately known, the PVT method can be extremely accurate, but when applied to commercial liquid propane and butane, whose compositions vary, the accuracy of the method is significantly reduced. For the high accuracies targeted it was essential to use densitometers.

The densitometers chosen were of a type incorporating a resonant cylinder. This type of instrument has a good record for reliability and can give an accuracy of about +/- 0.2 %, if an appropriate routine of servicing and calibration is pursued. A pair of such instruments per product (duplication to provide redundancy and allow averaging and cross-checking) were thus chosen for installation in the immediate vicinity of the flow meters.

It was originally hoped to install a second pair of densitometers in the storage tanks, for use in conjunction with the gauging system, but it was found impractical to do this. Instead it was decided to use PVT data to compute the density difference between the tanks and the densitometers near the flow meters, and then apply the appropriate small correction to the densitometer readings to determine the density in the tanks. Although PVT data were judged insufficiently accurate for the direct computation of density, error analysis showed that they were of adequate accuracy to compute such small corrections.


7.1 The original system

The original instrumentation planned for each tank is shown in Fig. 3; a conventional pressure gauge for pressure readings, a copper averaging thermometer to give an indication of the 'average' temperature within the tank, and one of two (one duty, one back-up) level gauges for level measurement. But for an orifice measurement of flow to the ships, there was no measure of any of the flows in to, or out of, the tank. To render the tanks suitable for proving the newly planned custody transfer flow meters a number of modifications to the tankage system were required to enable an accurate determination of the net mass of product moved out of a tank during a given interval of time.

7.2 The level measurements

Since the most accurate type of servo-operated level gauges were originally planned for the tanks, the only improvement to the accuracy of liquid level measurement that could be made was to average readings from as many level gauges as possible per tank. It was hoped to add a third to the two already being installed but circumstances prevented this. However, merely averaging the two provided a useful improvement.

The accuracy of liquid and vapour volume measurement, the determination of which was the real objective behind the improvement in liquid level measurement, was further improved by ensuring that the tanks were calibrated to the highest accuracy possible.

7.3 Pressure, temperature and density measurements

To convert tankage volume measurements to tankage mass measurements it is necessary to know pressure, temperature and density in both the liquid and vapour spaces of the tank. The existing pressure measuring devices were satisfactory for this purpose, but it was necessary to replace the original single copper resistance thermometer by a much more accurate temperature measuring system which could clearly identify the different temperatures in the tank's liquid and vapour spaces, and measure liquid stratification. The type of instrument chosen comprised sixteen platinum resistance thermometers spaced along a vertical line from top to bottom of the tank. The way in which the resulting pressure and temperature data are used to derive density values is then dealt with as broadly described in Section 6, above, but more particularly in terms of 16 horizontal (stratified) layers, each at the temperature as measured by one of the 16 thermometers.

7.4 Additional flow measurements

An essential requirement for determining the mass of product moved out of a tank during a given interval of time is to meter all flows both entering and leaving it. To reduce the number of meters necessary to do this it has been made an operational pre-condition that a tank lined up as a prover is neither supplied from the plant nor fed by the recirculation flow. Since the flow to the ship is already to be measured by the custody transfer meters, the necessary condition can be met by only additionally metering the vapour flows.

The desire not to modify the pressure profile in the lines around the vapour compressors, while deriving a moderately accurate flow reading over a fairly wide flow range, led to the choice of insertion type flow meters. The availability of excellent metering conditions for the instruments, straight pipe lengths well in excess of 100 pipe diameters, provided an opportunity for them to be used to their best advantage, but still some consideration had to be given to select the most suitable type.

Insertion turbine meters were not favoured because of their need for periodic calibration and uncertainty about their stability and durability in low temperature, non-lubricating vapour flows. Pitot tubes did not have the required range. Vortex meters, with he advantages of low blockage, wide range and inherent, known calibration, seemed quite attractive. Even intrinsically safe versions were available. Uncertainty as to whether their accuracy of +/- 5 % in an ideal situation would unduly effect the accuracy with which the loading line flow meters could be calibrated was dispelled by the error analysis referred to in Section 9. Accordingly they were chosen.


A central computer was needed to handle all the data to prove the delivery meters. It's role was to receive data from all the instruments in the system by way of a 15 second scan cycle, compute fluid movements in terms of mass, and repeatedly update stored data. On each scan it must check that readings are within expected limits and alarm if they are not.

The 15 second totalizations of mass are to be added into running totals started when ship loading begins. When it is expected that ship loading and all other operational conditions are to remain steady for a considerable time it will be possible to manually initiate a meter prove operation to cause the computer to record all instrument readings and running totals every 15 seconds. Another manual operation, perhaps two or three hours later, will terminate the proving operation and cause the computer to output the recorded readings and the flow meter's 'meter factors' based on the measurements in and around the tanks. A later study of this information will be made to confirm, or perhaps result in a change to, the flow meter's, meter factors.

During normal operation the central computer is to also be actively involved in the on-line determination of the ship's cargo. It will be able to give an accurate figure for the quantity of each product loaded, even if one custody transfer flow meter and both densitometers per product fail during loading. The failure of both densitometers on one loading line will cause the computer to use a pre-programmed density value. When a loading operation has been completed the computer will be able to print out directly from its memory a completed ship's loading certificate.

If the central computer fails, independent flow computers associated with each delivery meter should still give an accurate record of the amount of each product loaded on board ship, as measured by the custody transfer flow meters and densitometers. As a further back-up, a measurement of the amount of product loaded should always be available directly from the tank level gauges, although it will be of lower accuracy because it will have taken no account of vapour boil-off and will otherwise have a cargo sized dependent accuracy as described in Section 9. Ship loading capability should not be lost even if all the terminal's measuring and data processing equipment fails. Thus, even in this unlikely event, it should still be possible to load at least calibrated ships.


Equations have been developed which express, for different loading conditions, the estimated uncertainty in the measurement of mass delivered to a customer in terms of each individual instrument's measurement uncertainty.

In the space available it is not possible to give the derivation of these equations, but their nature can be illustrated by considering the equation for the percentage uncertainty, UD in the mass delivered from the storage tank down the loading line, as deduced from the upgraded tank gauging system, including subsidiary flow meters.

It should be noted that this is a condensed equation in which several of the symbols denote quantities that are themselves functions of other variables which need separate evaluation. In particular UT (the uncertainty in the tank gauge measurement system --- ignoring the effect of the secondary flows into and from the tank) is the result of a calculation involving measurements made on both the vapour and liquid phases in the tank.

expected to vary from about +/- 0.3 % for deliveries of 30,000 tons to over +/- 1.0 % for deliveries of only a few thousand tons taken from near the top of the tank.

Applied to the metering system described in this paper the error analysis :

(i) confirmed that the inherently less accurate but also less expensive and less expensively installed insertion flow meters were adequate for metering the vapour flows;

(ii) clarified the choice of where it was necessary to measure temperature and pressure in the vapour lines, and where temperature and pressure could safely be inferred from measurements elsewhere;

(iii) showed that the tank could be used to prove the 24 inch flow meters only when large deliveries at a reasonably constant flow rate were being made.

(iv) showed that for small deliveries the meters could be expected to be considerably more accurate than the tank gauging system.


This project has opened up a new area in metering. For what is believed to be the first time a liquefied gas storage tank has been instrumented in such a way that, when delivery conditions are suitable, it can be used to prove a flow meter. By linking the tank with 24 inch ultrasonic and vortex shedding delivery meters --- both of which are known to have good long term stability, so that they need only infrequent proving --- a metering system for refrigerated N.G.L. has been designed which should be of adequate accuracy for custody transfer purposes. Thus the large scale custody transfer metering of refrigerated hydrocarbons, including even N.G.L., now appears to have become a practical possibility.


The authors wish to thank Abu Dhabi Gas Industries Ltd., for permission to publish this paper.


1. Inkley, F.A., Walden, D.C. and Scott, D.J. : "Flow characteristics of vortex shedding flow meters", Measurement and Control 13, May 1980, p.p. 166-170.