Flow Measurement and Instrumentation 17 (2006) 141–152

www.elsevier.com/locate/flowmeasinst

Development of weighing tank system employing rotating

double wing diverter
R. Doihara ∗ , T. Shimada, Y. Terao, M. Takamoto
National Metrology Institute of Japan (NMIJ), AIST, 1-1-1 Umezono, Tsukuba, Ibaraki 305-8563, Japan

Abstract

A new design for a diverter and weighing tank system has been developed. Experiments have been carried out on a prototype. Diverter wings
are set as partitions along the diameter of a half-cylinder tube. The rotation axis is at the center of the half cylinder and runs parallel to liquid jets
flowing vertically downward from several independent nozzles. This design contributes to small diverter timing error, compact design, small wet
area, a simple and robust mechanism, easy implementation of the ISO4185 test, and easy adjustment of diverter trigger timing. After adjustment
of start and stop signal timings, the estimated standard uncertainty caused by diverter timing error is reduced within a negligible small value
(0.0019% or less) over a wide range of flow rates at any nozzle position. The weighing tank system has been improved by some refined devices.
A single movement of the weighing platform can prevent escaping vapor, enable connection of an air tube, provide protection against impact, and
provide a fail-safe against overflow. It is confirmed that the liquid flow calibration facility using the present system achieves accurate measurement,
easy operation, wide flow range, high throughput, low cost, small weighing tank, easy maintenance, and safety.
c 2006 Elsevier Ltd. All rights reserved.

Keywords: Diverter; Weighing tank system; Double wing; Liquid flowmeter; Calibration

1. Introduction calibration throughput is low. For a wide-ranging calibration

The calibration of a flow meter consists of the comparison of
the output of that flow meter with a standard. There are various
methods to establish a liquid flow rate standard [1]. Systems
using volumetric tanks, pipe provers or compact provers are
volumetric methods. A weighing tank system is one of the
gravimetric methods. A static weighing tank system employing
a flying start and stop method has excellent features. In fact, a
weighing tank system with a diverter has been adopted in many
calibration facilities [2–7]. In this system, it is advantageous
that the weighing scale can be calibrated by dead weights,
which are extremely stable. In addition, the diverter does not
disturb the flow condition in the test line when it switches the
flow line (to a bypass course or to a collection tank) at the start
and end of measurement.
However, there are some weaknesses. The calibration cost
tends to be high, because the construction is expensive and the
∗ Corresponding author. Tel.: +81 29 861 3816; fax: +81 29 861 4242.
E-mail address: r-doihara@aist.go.jp (R. Doihara).
service, many sets of weighing tanks and diverters have to be
made, because each calibration line has a limited flow range. A
relatively long time is needed for each test in order to reduce the
influence of diverter timing errors. Essentially, at the inlet of the
weighing tank, an open section exists in order to mechanically
disconnect the weighing scale from the pipelines. In the case
of oil flow rate standard facilities, the diffusion of oil vapor,
or mist, to the atmosphere is unfavorable. Also, for a water
flow facility, evaporation from this open section can be one
of the error sources. There is also the possibility of overflow
at the time of system malfunction. Some connections to the
weighing tank can be sources of error. In many cases, a power
tube is connected to a drain valve for automatic operation.
Additionally, signal wires for liquid level sensors are linked to
the tank.
In particular, the diverter timing error is a significant
problem. The problem arises at the start and end of the diversion
period. If the diverter is not set to trigger the timer at the correct
instant, significant error time can be introduced into the time
measurement. This can be a major source of uncertainty at

0955-5986/$ - see front matter 

c 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.flowmeasinst.2006.01.002
142 R. Doihara et al. / Flow Measurement and Instrumentation 17 (2006) 141–152
maximum flow rate, because the contribution of the diverter
timing error increases as the collection time decreases.
ISO4185 [8] recommends adjustment of the trigger timing
on the basis of the diverter error time obtained from estimation
tests. In the traditional diverter, the correct timing exists at only
one symmetric point against a liquid jet velocity profile [9,10].
Confirmation and adjustment of the error should be performed
under all calibration conditions, because the liquid jet velocity
profile is dependent on the flow rate and the shape of the nozzle.
However, estimating the error under all calibration conditions is
very difficult, resulting in significant uncertainty derived from
the diverter timing error.
Traditionally, several techniques for reducing the diverter
error have commonly been used. A larger collection tank
has been adopted to obtain a longer collection time. Trigger
actuation has been set to be as symmetric as possible. The
diverting wing has been made to move at high speed to reduce
the transit time of the liquid jet, resulting in smaller diverter
timing error. However, splashes can occur when the speed is
too fast.
Various new attempts have been made to reduce this error.
At PTB, the nozzle and the feeding pipe have been specially
designed by computational fluid dynmics (CFD) simulations
to achieve a symmetrical velocity profile at the liquid jet [11].
Furthermore, a new method for estimating timing error by using
an angular encoder and storage device was proposed to enable
easy adjustment of the trigger position [12].
Recently, NMIJ developed a double-wing diverter [13]
and succeeded in minimizing the timing error. A similar
unidirectional motion diverter was also designed by NIST
[14,15]. This new system has two diverter wings which, at
the start and end of measurement, are made to move in the
same direction at a constant speed across the liquid jet. After
appropriate adjustment, the double-wing diverter maintains a
small diverter timing error for any given flow rate and any
given nozzle condition. This system is already used in the
primary standard facility for NMIJ [6]. Uncertainty analysis
has confirmed that the contribution of the adjusted diverter
timing error is negligible compared with other sources of error
[16]. However, the double-wing diverter has a problem in the
implementation of the estimation test under ISO4185. Details
of the problem and its solutions are discussed in the subsequent
sections.
In the near future, NMIJ plans to extend the facility’s range
to smaller flow rates by providing another calibration line.
Therefore, the authors have developed an improved double-
wing diverter and an advanced weighing tank system to resolve
the problems mentioned above in the construction of the small
test line. The new design of this system aims to satisfy several
demands: accurate measurement, easy operation, a wide range
of flow rate, high throughput, low cost, a small weighing tank,
easy maintenance, and safety.
In the present paper, the design of the diverter and weighing
tank systems is proposed. The diverter timing error was
estimated using a small prototype in a water flow circuit in order
to compare the features of the new diverter system with the
previous system, and to confirm that the error time was reduced
to be negligibly small. Some demonstrations have been carried
out on the prototype in order to confirm that the new system has
the features demanded.
2. Development of rotating double-wing diverter
2.1. Diverter timing error
On the basis of the static gravimetric method, the standard
mass flow rate q is given by Eq. (1).
m1 − m0
q= (1)
tC (1 − ρA /ρW )
tC = tTRUE − t. (2)
Here, m 0 is initial weight, m 1 is weight after collection, ρ A
and ρ L denote the density of air and liquid, respectively, and tC
is the time interval between the starting and stopping signals.
The diversion causes a timing error in the measurement of
collection time. The collection time tC may include the diverter
timing error t. This timing error influences the calibration
results systematically, and cannot be reduced by repeating
calibration. The diverter system is thus a key component in
achieving a highly accurate liquid flow rate standard.
Fig. 1 shows typical time profiles of the flow rate directed
to the weighing tank. Ideally, the diverter should change the
flow direction in a moment of time. But that is impossible
in practice. Some descriptions of the flow rate, qIN , into the
weighing tank are shown in the transient flow rate profiles in
Fig. 1. The tTRUE value is the time equivalent to the collection
time assuming the diverter is ideal. In the case of a conventional
single-wing diverter (SW Diverter), the ideal trigger timing
may move when the velocity profile of a liquid jet, which is
discharged from the nozzle, changes. Using this diverter, the
flow rate profile becomes symmetric, as shown in Fig. 1(a)
because the single wing moves across the path of the liquid
jet in opposite directions at the start and end of measurement.
Thus the collection time should be measured at an exact timing
condition (S1 = S2 and S3 = S4) in order to obtain the ideal
collection time tTRUE [9,10]. If the trigger timing does not
satisfy the exact condition, the timing should be adjusted to the
exact timing, which exists at only one point. When the flow rate
profile changes because of a different flow rate or a different
nozzle width, the true timing tTRUE can be of different period,
such as in Fig. 1(b). In the case of Fig. 1(b), the collection time
tC contains the diverter timing error t, because the timing
described by the dotted line does not satisfy the condition
(S1 = S2 and S3 = S4). Commonly, it is said that the timing
error should be less than 10 ms at a condition of maximum flow
rate. Otherwise, the collection time needs to be longer than 100
s to make the contribution of the timing error to the calibration
uncertainty less than 0.01%. Therefore, a diverter system needs
a large weighing tank or it limits its maximum flow rate.
In view of this problem, the double-wing diverter (DW
Diverter) was developed by Shimada et al. [13,17]. The new
system has two diverter wings which, at the start and end of
measurement, are made to move in the same direction at a
constant speed across the liquid jet, as shown in Fig. 2. In
 R. Doihara et al. / Flow Measurement and Instrumentation 17 (2006) 141–152 143

Fig. 2. Schematic of DW diverter.

flow rates, as long as the profile remains stable from the start
to the end of the diversion. The value of diverter timing is a
function of the difference between the distance of the wings
L wing and the distance of the sensors L sensors when the moving
speed is constant. The condition (S1 = S3 and S2 = S4,
thus tC = tTRUE ) can exist for any timing as long as L wing =
L sensors, and the moving speed is constant. See the reference
article [17] for the detailed theory of the double wing method.
The condition (S1 = S3 and S2 = S4) is maintained at any flow
rate profile after adjustment. Therefore, once the diverter has
been adjusted appropriately, adjustment of the trigger timing
is unnecessary for any given flow rate and any given nozzle
position.
However, the double-wing diverter described above has
practical difficulties in the implementation of the ISO4185 test,
because the two wings are moved in the same direction.

2.2. Implementation of estimation test in ISO4185

The diverter timing error should be estimated experimen-

tally. The following test (a series of short flows) is one of
the estimating methods recommended in ISO4185. First, the
initial weight m 0 is measured. Then the weighing tank is filled
up by a single diversion to measure the weight after collection
Fig. 1. Flow rate transition profile and trigger timing. m 1 andthe duration of diversion tC1 . Next, the total weighing
scale ni=1 m i and total time ni=1 ti are measured in a series
the DW diverter, the diverter timing error is not dependent on of n (as many as 25) diversions without resetting the timer and
the flow velocity profile of the liquid jet from the nozzle. For the scale. Finally, the error results are obtained from these val-
the DW diverter, if the liquid jet has asymmetric velocity, the ues by the following formula:
⎛  n ⎞
flow rate adopts the profile shown in Fig. 1(c). Therefore, the n 
mi ti
condition that the diverter timing error becomes zero is S1 = S3 tC1 ⎜⎜ q1 i=1 i=1
⎟
⎟
and S2 = S4. In practice, trigger signals are generated by two t = ⎜ · − 1⎟ . (3)
n − 1 ⎝ qn (m 1 − m 0 )/tC1 ⎠
photosensors which detect a pin on the moving unit. Thus, each
trigger can be adjusted independently by the sensor positions.
Even if the first trigger is generated at any timing, the diverter Here, q1 represents the flow rate during a single diversion
timing error is able to be reduced, as long as the second sensor measured by a flowmeter, and qn is the average of the flow rates
is adjusted to the condition (S1 = S3 and S2 = S4). Even if measured by the flowmeter during n diversions.
the velocity profile of the jet changes as the flow rate changes, If the duration of diversion is sufficiently long in comparison
as shown in Fig. 1(d), the error time is almost constant over all with the timing error, the error will be negligible. However,
144 R. Doihara et al. / Flow Measurement and Instrumentation 17 (2006) 141–152
in many calibration facilities for large flow rates, the duration
is insufficient because of economical or physical limitations.
In many calibration facilities, the error caused by the diverter
has often been the main source of uncertainty. This error also
constitutes one reason to limit the maximum flow rate with
reference to the volume of the weighing tank.
Even if a diverter is designed on the principle of the double-
wing method, some tests for estimating diverter timing error
must be conducted for adjustment of the timing. In the previous
DW diverter shown in Fig. 2, the end position is different from
the start position because the wings travel linearly in the same
direction. When the wings must return to the start position
for the next diversion, the liquid jet drops into the weighing
tank because the wings pass under the liquid jet. In the test
under ISO4185, this mechanism leads to a problem in practical
operation.
There are some solutions to prevent this unnecessary inflow.
One method is to install an additional diverter at the duct
between the main diverter and the weighing tank. But some
liquid that is not intended to drop into the weighing tank
may fall and wet the wall of the collection side. This error
mechanism can become large in the case of a small oil facility,
because the wet liquid on the wall becomes significant relative
to the small resolution of the weighing scale. A large wet
area also lengthens the waiting time for the liquid to run
down completely. In the case of oil, the wet area should be
made as small as possible, since the waiting time is apt to
increase drastically because of high viscosity. Determining a
method for checking for leaks presents another problem. Using
this additional device only for the estimation test will reduce
accuracy in usual calibration; in addition, ISO4185 tests are
seldom carried out.
Another method is to set a plastic bag between the diverter
wings in order to prevent the unnecessary inflow. However, this
involves setting the bag by hand at each diversion. In an oil
facility, especially, this operation is terribly hard and unsafe.
2.3. Design of the rotating double-wing diverter
Using a new design, we have achieved small diverter timing
error, compact design, small wet area, a simple and robust
mechanism, easy implementation of an ISO4185 test, and easy
adjustment of diverter trigger timing. This diverter system is
based on the concept of symmetric diversion by two wings that
move across the liquid jet in the same direction at constant
speed at both the start and end of measurement. In the present
design, the wings are rotated, whereas in the previous double-
wing method they are moved linearly.
Fig. 3 illustrates the design of the diverter with two rotating
wings. The diverter wings are set as partitions along the
diameter of a half-cylinder tube. The axis of rotation is at the
center of the cylinder and runs parallel to the liquid jets that flow
vertically downward. One side of the diverter has a rectification
surface that resembles a half cone with a closed bottom (Bypass
side). When this side comes under the nozzle, the liquid flow
is led into a bypass line. The other side (Collection side) is a
semicircle shape with an open bottom, such that the liquid jet
can drop directly into the weighing tank. The diverter is rotated
by a servo motor via a center shaft linked to a timing belt. Two
photosensors and a disk with a slit are used as a trigger signal
generator for starting and stopping the timer, and these are set
at the top of the rotating shaft.
Before collection, the half cylinder is set in the bypass
position to locate the bypass side under the nozzle. In this paper,
this angular position is defined as 0◦ . To start collection, the
cylinder is rotated 180◦ (to the collection position). On the way
to the collection position, at 90◦ , one of the diverter wings cuts
the liquid jet. At that moment, one of the photosensors detects
the slit of the disk and sends the start trigger signal to the
timer. While the cylinder is in the collection position, the liquid
jet falls into the weighing tank. When the tank has filled, the
cylinder again rotates 180◦ in the same direction (to 360◦ ) in
order to stop collection. The other diverter wing cuts the liquid
jet at 270◦, and the other photosensor sends the stop trigger
signal to the timer. The sensors are fixed on a precision traverse
stage so that the sensor position can be adjusted on the basis of
the estimated diverter timing error.
In contrast to the previous diverter, the new design offers
many advantages: the wet area can be minimized; the diverter
wings do not have to return, because the end position is the
same as the start position; the wet area of the collection side
is completely separated from the bypass side; the ISO4185
test can be run consecutively without any waiting time; usual
calibration can be carried out accurately; and the mechanism of
this rotating motion is simple and robust. However, tests should
be carried out in order to confirm that the rotating motion of the
diverter poses no problem.
3. Improvement of weighing tank system
In many cases, weighing tank systems have an open section
at the inlet, through which vapor might escape from the
collected mass. A weighing tank is often linked by electric
wires to detect the liquid level, and by air tubes to actuate the
drain valve. These can be sources of error. Especially in the case
of a weighing tank with a small scale, even if the tension caused
by those connections is small, these errors could exert a greater
influence on measurement than the resolution of the scale.
The weighing tank that we have used includes devices to
measure the mass of collected liquid accurately. The weighing
tank in the present system has no open section from which
vapor and mist can escape into the atmosphere, except for
a short time during weight measurement. Fig. 4 shows the
sequence of operation of the weighing system. The weighing
tank is placed on a lift-platform which is elevated by air
actuators. The tank is lifted from the scale and pushed in tightly
against the bottom of the diverter box, in order to protect
the scale against sudden impact by the jet and to prevent the
escape of vapor. Although the air inside the tank flows from the
weighing tank to the diverter box as the tank is filled with liquid,
the air inside the box has already been saturated with mist or
vapor. Therefore, diffusion and evaporation hardly progress.
Even in the event of unexpected overflow due to malfunction
of the diverter, the liquid will flow to the bypass line through
 R. Doihara et al. / Flow Measurement and Instrumentation 17 (2006) 141–152 145

Fig. 3. Design of diverter with rotating double wing.

Fig. 4. Sequence of operation for the weighing system.

the diverter box. This tight sealing at the inlet functions as a
fail-safe mechanism.
The weighing tank is placed on the scale in a completely
free state during weighing, because no electric wires or air
tubes are connected to the weighing tank. In order to avoid any
mechanical connection to the tank, the liquid level is detected
by optical sensors and a float inside a transparent tube. The
automatic operating system can discriminate the liquid level of
the weighing tank by detecting the float with the optical sensor.
When the optical sensor that is set at the desired full level mark
detects the float, the diverter is rotated. If an optical sensor
set at the very top of the tank detects the float, the system is
immediately shut down to ensure safety in an emergency. An
optical sensor set at the lowest position closes the drain valve
after a sufficient interval after the time the float passes by the
sensor.
The system also features automatic connection of an air tube
for the drain valve. When the tank is pushed up, the connection
is established, allowing the valve to be operated by the air
signal. When the tank is placed on the scale, the connection is
completely separated. These devices are especially adaptable
and useful in a hydrocarbon flow facility. In such a facility,
oil vapor ejection to the atmosphere is unfavorable in view of
measurement accuracy and the environment of the room. The
air connection is safe in a hazardous area. Of course, in the case
of an oil facility, the necessary ground wire to be connected
must be as flexible as possible.
146 R. Doihara et al. / Flow Measurement and Instrumentation 17 (2006) 141–152
4. Performance test of the new diverter and improvement
of other components of the calibration system
4.1. Application to a water flow facility
Fig. 5 shows a schematic of the experimental apparatus and
flow circuit. The inner diameter of the main piping is about
25 mm. The flow rate is maintained from 10 kg/h to 3000
kg/h by means of a pump, a control valve and a bypass valve.
The water is automatically maintained within 24.1 ± 0.1 ◦ C
by a heat exchanger. The maximum weighing range of the
scale is 32 kg, with a resolution of 0.1 g. The weight of the
weighing tank is near to 12 kg. Thus the weighing system has
a capacity of 20 kg for the collected liquid. The weighing scale
can be calibrated using six 5 kg dead weights. Atmospheric
pressure and ambient temperature are measured for buoyancy
correction. The ambient room temperature is also constant
at about 24 ◦ C. Three Coriolis flowmeters are used as test
meters for calibration. One timer for the diverter, and two
counters and two timers for each flowmeter, are equipped to
calibrate the two flowmeters simultaneously. In order to use the
double chronometry pulse interpolation technique, a gate signal
generation circuit has been made with a flip-flop logic IC.
4.2. Small storage tank with overflow system
For a weighing tank system without an over flow head
tank, the flow rate may become slightly reduced because the
liquid level over the pump suction in the storage tank decreases
while the liquid is being collected in the weighing tank. Many
water flow facilities have a head tank system. However, such a
system is difficult in the cases of an oil facility or high pressure
loss flowmeter. The change in the liquid level is especially
significant in the case of a small storage tank. As such, several
solutions have been proposed. In one system, an extra buffer
tank is installed with enough liquid to supply the storage tank
during collection in order to maintain the liquid level [6].
This prototype facility has a small storage tank (0.56 m3 )
and has been modified to counter the storage tank problem
described above. Before modification, the flow rate was reduced
by about 0.1% or more due to a liquid level decrease of about 25
mm as the liquid accumulated in the weighing tank (0.02 m3 ).
For the modification, an overflow system has been established
inside the storage tank by using two pumps. The inside of
the storage tank is divided by a partition board into two tanks
(suction side and buffer side). One main pump maintains the
flow rate to the test line from the suction side. The other pumps
liquid to the suction side from the buffer side so that liquid
always flows over the partition board. As a result, the liquid
level is almost constant (within 2 mm) during collection.
4.3. Multi-nozzle system combined with RDW diverter
Downstream from the control valve, the piping branches
into four nozzles which feed water jets to the diverter in the
prototype. The inner diameters of the nozzles are 18, 13, 7, and
4 mm, respectively. Many calibration facilities equipped with
Fig. 5. Schematic diagram of the prototype.
a diverter adopt a variable-width nozzle in order to maintain
a stable jet flow over a wide range of flow rates, because
one diverter must be fed by one nozzle [4–6]. Thus, in most
facilities using a diverter, the test line pipes are all combined
together to fit the one nozzle size that gives the maximum
flow rate, although such facilities have several test lines with
different sizes of pipe. For small test lines, this confluence is
unfavorable, because the dead volume between the test meter
and weighing tank becomes large and the flow velocity is
reduced greatly. When the flow rate is low, removing bubbles
from the top of the variable-width nozzle is difficult. In
addition, such a variable device has a complicated mechanism
and may be prohibitively expensive, especially in the case of a
small-sized system.
In our prototype, independent multi-nozzles of varying inner
diameters are employed. Therefore, the dead volume can be
minimized by selecting the nozzle according to the flow rate.
The flow velocity at the top of the nozzle is maintained at a high
rate even at a low flow rate. These nozzles are nothing more
than a cutting edge of the pipe. According to the theory of the
double-wing method, the diverter timing error is not dependent
on the position and flow velocity profile of the liquid jet from
the nozzle [17].
However, it should also be confirmed experimentally that the
multi-nozzles system poses no problems.
5. Results of performance test and discussion
5.1. Performance of the RDW diverter with multi-nozzle system
5.1.1. Comparison between RDW diverter and SW diverter
In our experiments, the diverter timing error was estimated
by Eq. (2) of the ISO4185 test. We set the diversion number at
n = 25. The flow rate was measured by a Coriolis flowmeter.
The angular velocity of the wing was set to 360◦/s. In some
experiments simulating a single wing diverter, only one wing
was used to switch the flow. For this one wing simulation, a
symmetrical-type wing was adopted which has 3◦ angle edges
at both sides of the tip and which is 6 mm thick. In order to
 R. Doihara et al. / Flow Measurement and Instrumentation 17 (2006) 141–152
(a) Nozzle position: A.
Fig. 6. Position of multi-nozzles.
Fig. 7. Estimated diverter timing error before and after adjustment for single
and double wings.
investigate the influence of the flow rate profile and the nozzle
position, two nozzle positions were tested: a straight position
(A) and an oblique position (B), as shown in Fig. 6. In practical
calibration, the valves of the larger nozzles are closed as the
flow rate becomes small. In these experiments, all four valves
were always open.
The results of the estimation test of the diverter timing
error are shown in Fig. 7. Nozzle position A was used in this
experiment. First, the photosensor for the start signal was set
at 90◦ . The diverter timing error was measured using only one
wing in order to simulate a single-wing diverter. The averaged
diverter timing error for the single wing was about 7.5 ms,
because the nozzles are not set exactly aligned with the 90◦
line (Fig. 6). Second, from the time of error and the angular
velocity of the wing, the adjustment angle of the photosensor
was calculated, corresponding to −1.4◦. After the photosensor
position was adjusted to near 88.6◦, the estimation tests were
again carried out using the single-wing method. Next, with the
position of the photosensor maintained near 88.6◦, the other
photosensor was set at 270◦. Then the diverter timing error
was measured by the double-wing method. The average error
time and correction angle for the double wing were −2.1 ms
and −0.8◦ , respectively. Finally, after the other sensor position
was adjusted to 269.2◦, estimation of the timing error was
again carried out with two wings. After adjustment of the
sensors, the diverter timing error fell within 1 ms for the double-
wing method regardless of flow rate. The averaged value of
the diverter timing error after adjustment t was −0.1 ms.
147
(b) Nozzle position: B.
Table 1
Average of the diverter timing error t and its standard deviation σt
Nozzle t (ms) σt (ms)
A −0.8 0.6
SW diverter (Symmetric wing)
B −37.9 2.6
A −0.1 0.4
RDW diverter (Symmetric wing)
B −0.4 0.4
RDW diverter (Asymmetric wing) A 0.1 0.2
The standard deviation σt was 0.4 ms. These values are
summarized in Table 1.
The results of the single wing method suggest that the
diverter timing error is somewhat dependent on flow rate. The
flow rate dependency of the timing error was 1.5 ms at the
maximum difference among averaged values on each flow rate.
This value was less than the value that was expected. But it
depends on the liquid jet condition, which is namely the nozzle
ability. This value might be small, because the nozzles consist
of a plural number of circular pipes that have relatively small
diameter. The important point is that the DW diverter has no
flow rate dependency, although the SW diverter can be affected
by flow rate. Especially in the case of this multi-nozzle system,
a combination of open nozzles will change to keep the flow
velocity high. The center position of the liquid jet is easily
moved, because these nozzles are not set exactly aligned along
the straight line. Therefore, we discuss how the RDW diverter
is able to maintain the smallest diverter timing error versus
different liquid jet conditions.
5.1.2. Performance at different nozzle positions
Diverted flow rates were measured versus the angle of the
diverter wing as follows. The diverter wing enters the region
of the liquid jet and is stopped at an arbitrary angle. During
a constant time (30 s), a portion of the liquid jet flows into
the weighing tank. The flow rates are then calculated from the
weight of liquid and the time. The angles are set from 80◦ to
130◦ in increments of 0.5◦ .
Fig. 8 shows the measurement result of the diverted flow rate
(1500 kg/h) versus diverter wing angle. The flow rate pattern
for position A is relatively symmetric. The pattern for position
B, however, is quite bumpy and asymmetric. In the case of
the single-wing diverter, the diverter timing error theoretically
becomes zero [9,10] under the conditions of S1 = S2 and S3 =
S4, as illustrated in Fig. 8. The angle that satisfies S1 = S2
was calculated from the flow rate data under the premise that
the velocity of rotation is constant. The calculated angle for the
148 R. Doihara et al. / Flow Measurement and Instrumentation 17 (2006) 141–152
Fig. 8. Diverted flow rate versus position of the diverter wing.
Fig. 9. Diverter timing error at different nozzle positions.
position A is 88.8◦, which is close to the adjusted angle in the
single-wing method. The angle for position B was 95.6◦.
Fig. 9 shows the results for position B. The averaged value
of the double-wing method was −0.4 ms. The RDW diverter
has succeeded in keeping small timing error. Although the jet
condition was changed remarkably, the results confirm that,
in the case of the double-wing method, the diverter timing
error is not dependent on the flow profile. A conspicuous
difference from the single-wing method is observed. The value
of the diverter timing error at 1500 kg/h is −39.8 ms, which
corresponds to an angle of 95.8◦ for the sensor position.
Obtaining such close values using different methods indicates
the high reliability of these experiments.
According to these experiments, it is confirmed that
the rotation motion of the double-wing and multi-nozzle
system does not cause problems with calibration. Therefore
these systems have the following advantages: they can
maintain accuracy using the double-wing method, allow
easy implementation of ISO4185 test, operate with a simple
mechanism at low cost, and prevent the persistence of bubbles.
Fig. 10. Adjustment of diverter timing error cased by wing with reduced
splashing.
5.2. Adjustment of diverter timing error
In order to make final adjustments, the wings were replaced
with by an asymmetrical type that has one side edge at 3◦ to
reduce splashing. The optimum shape of a wing cross section
has already been investigated [13]. It is desirable that the entry
side is flat, and that the wing should be as thin as possible. The
motion is also important, in that the flat plane should always be
kept parallel to the liquid jet. The test meter was changed from
the previous experiments to a larger-size Coriolis flowmeter.
The sensors positions were reset. The diverter timing error
was measured again to estimate the readjustment value. Fig. 10
shows the diverter timing errors. Before the readjustment, the
averaged value of timing error was t = −4.4 ms. Then,
after readjustment, it is confirmed that the averaged diverter
timing error and the standard deviation were t = 0.1 ms and
σt = 0.2 ms, respectively.
5.3. Demonstration of rapid calibration
As a result of calibration, we can obtain the K -factor of
a flowmeter under calibration. If the diverter timing error is
insufficiently small compared with the collection time, the
K -factor is greatly affected as the collection time decreases. In
a calibration facility using a flying start and stop method, the
K -factor is calculated from the pulse frequency of the
flowmeter and the measured standard mass flow rate:
fp
K = (4)
q
where K is the K -factor, and f p is the pulse frequency of the
flowmeter. The affected K -factor resulting from the diverter
timing error is derived from Eqs. (1), (2) and (4):
fp (1 − ρA /ρW )
KD = = fp (tTRUE − t)
q m1 − m0
t ∼ t
= K TRUE 1 − = TRUE
K 1 −
tTRUE tC
(5)
 R. Doihara et al. / Flow Measurement and Instrumentation 17 (2006) 141–152 149

Although the standard uncertainty of the indicating time of

the timer u(tCm ) corresponds to the resolution of the timer
(1 × 10−6 s), it is small enough to be ignored. The timer
was calibrated with a standard timer. As a result, the relative
standard uncertainty u(tCC ) of the correction time for timer
calibration is estimated to be 1.3 × 10−5 (−). The uncertainty
attributed to the reproducibility of the timer depends on the
stability of the built-in quartz oscillator. By referring to the
statement given in the timer’s specifications, it is assumed that
the oscillator features an accuracy of 2.5 × 10−5 (−) or less and
that this accuracy value is equal to the limit of error in a triangle
distribution. Based on a reference [18] GUM 4.3.9 Eq. (9b), the
relative standard uncertainty is estimated as

u(tCδ ) 2.5 × 10−5

= √ = 1.0 × 10−5 . (8)
tC 6
The basic equation for analyzing the uncertainty in timing
Fig. 11. Demonstration of short time calibration.
error time is given by Eq. (9):
where K D is the affected K -factor resulting from the diverter t = t + tδ + tJet . (9)
timing error, and K TRUE is the non-affected K -factor. Thus the
affected K -factor K D should be proportional to the inverse of Here, tδ is the deviation from the average timing error, and
the collection time tC . tJet is the change in the error time caused by different liquid
Fig. 11 shows the calibrated K -factor when the collection jet conditions. The standard uncertainty in diverter timing error
time is intentionally shortened. All calibrations were carried is given by Eq. (10):
out at a constant flow rate (1800 kg/h). The abscissa presents 
the collection time and the ordinate indicates the K -factor.   2
u(t) = u t + {u (tδ )}2 + {u (tJet )}2 . (10)
Each symbol shows the average K -factor of 10 calibration
results. The error bars show the standard deviation of the 10 Average diverter timing error is t = 0.1 ms after final
calibration results for each collection time. The average K - adjustment. Its standard deviation is 0.2 ms. The anticipated
factor after the adjustment does not change, in spite of the short value of the average diverter timing error t is zero, and a
collection time. It is confirmed that the diverter timing error rectangular distribution is expected. According to GUM 4.3.7
after appropriate adjustment is negligibly small. The square Eq. (7) [18], therefore, the systematic effect of the timing error
symbols show the results when a significant amount of timing is estimated as
error, t = −4.4 ms, is observed. This value is taken from the
0.0001
estimated timing error before adjustment, as shown in Fig. 10. u(t) = √ = 6 × 10−5 (s). (11)
Even at the longest condition (40 s), a systematic shift of 3
0.012% was observed. There is significant error in the K factor
The uncertainty in the deviation from average timing error
at short collection times. The solid line shows the theoretical
corresponds to the uncertainty in the experimental standard
values calculated from Eq. (5), assuming that the K TRUE is
deviation of the mean:
999.87 P/kg and the diverter timing error t is −4.4 ms. It
is obvious that the effects arise from the diverter timing error. 0.0002
u(tδ ) = √ = 5 × 10−5 (s). (12)
Here, the uncertainty analysis of a measured collection time 20
is discussed to consider the contribution of the diverter timing
error to the overall uncertainty. The basic equation for analyzing The change in the error time tJet is 0.3 ms by comparison
the uncertainty of the collection time tC is given in Eq. (6): between nozzle positions A and B. A rectangular distribution
is also expected. The standard uncertainty in the change of the
tC = tCm + tCC + tCδ + t, (6) error time caused by different liquid jet conditions is estimated
where tCm , tCC , and tCδ are the indicating time of the timer, as
the correction time for timer calibration, and the reproducibility 0.0003
u(tJet ) = √ = 1.7 × 10−4 (s). (13)
of the timer, respectively. The standard uncertainty of the 3
collection time is estimated as
Based on Eqs. (10)–(13), the standard uncertainty in diverter
u(tC )
timing error is estimated as
tC 

    u(t) = 0.000062 + 0.000052 + 0.000172 = 0.00019 (s).
u(tCm ) 2 u(tCC ) 2 u(tCδ ) 2 u(t) 2
= + + + . (7) (14)
tC tC tC tC
150 R. Doihara et al. / Flow Measurement and Instrumentation 17 (2006) 141–152

Based on Eqs. (7), (8) and (14), the relative combined standard
uncertainty in collection time is obtained in Eq. (15):
u(tC )
tC

2
2
1 × 10−6  2  2 1.9 × 10−4
= + 1.3 × 10−5 + 1.0 × 10−5 +
tC tC


 3.6 × 10−8
= 2.6 × 10−10 + . (15)
tC2
Collection time tC (s) is approximately obtained from the
collected mass m 1 − m 0 (kg) and flow rate q(kg/s): tC ≈
(m 1 − m 0 )/q. Assuming that the collected liquid is 20 kg, Eq.
(15) is converted to Eq. (16):

u(tC )
= 2.6 × 10−10 + 9 × 10−11 q 2 . (16)
tC
When the collection time tC is quite short, for instance 10 s,
the contribution of diverter timing error becomes just 0.0019%
at any nozzle condition, which is a comparable value to the
uncertainty in the timer itself (0.0016%). Normally, the diverter
timing error is negligible in the case of the double-wing diverter
combined with a multi-nozzle system. Therefore, the RDW
diverter enables calibration at very high flow rates relative to
the weighing tank capacity.
5.4. Effect of the lift-up closure mechanism of the tank lid
In order to investigate the effectiveness of the lift-up
mechanism which seals the inlet of the weighing tank, the
weight is measured after the weighing tank is filled. Three
conditions are examined. The first condition is with the tank
raised and tightly sealed. The second condition is with the walls
of the diverter box removed to allow vapor to diffuse through
the collection side duct of the diverter. The third condition is
with the tank lowered to allow vapor to escape from the inlet of
the tank into the atmosphere directly.
Fig. 12 shows the weight loss from the filled weighing tank
with time. In the third condition, in which the inlet remained
open, the collected mass was lost at a rate of about 0.13 g/min
because of the evaporation and diffusion of vapor. In the second
condition, the loss rate became relatively small, because the
duct out of which the vapor could diffuse was narrow compared
with the third condition. The first condition demonstrates that
the tight seal of the weighing tank and the closed diverter box
succeeded in preventing evaporation loss. Therefore, even if the
weighing tank is filled at a low flow rate, thereby requiring a
long time, the evaporation effect will be negligible, as long as
the lift platform is in the raised position.
5.5. Test calibration at low flow rate
In this section, the effect of evaporation on the calibration
factor is discussed. A calibrated K -factor of the test meter is
affected by the flow rate of evaporation:
fp
KV = (17)
q − qV
Fig. 12. Change in the weight of filled weighing tank over time.
where K V is the affected K -factor of the test meter, and qV
is the evaporation flow rate. When qV << q, Eq. (17) is
approximated as Eq. (18):
fp 1 qV
KV = · ≈ K TRUE 1 + . (18)
q (1 − qV /q) q
Thus the calibrated K -factor increases with increasing
evaporation flow rate qV . It is obvious that the effect is
larger at low flow rates from the proportionality coefficient of
1/q. In order to evaluate the evaporation effect, calibrations
at low flow rates (20 kg/h) were carried out for the three
conditions described in the previous section. The test meter was
replaced by the small-size Coriolis flowmeter (max flow rate:
83 kg/h). Only one nozzle (4 mm) was used to maintain the
flow velocity of the liquid jet. In Fig. 13, the K -factors are
plotted assuming that the evaporation flow rates were the same
as the results of Section 5.4. As predicted by Eq. (18), the K -
factor is overestimated for the condition in which the tank was
in the lowered position. The least squares fit indicates that the
unaffected K -factor should be 1000.6 P/kg. This result is close
to the K -factor for the condition in which the tank is raised and
sealed tightly. Therefore, it is confirmed that the tight sealing
tank system effectively functions at a low flow rate.
The standard uncertainty in the contribution from liquid
vapor escape (m) is estimated as below:
m1 − m0
m = qV · tC = qV · (19)
q
u(m) u(qV )
≈ . (20)
m1 − m0 q
Assuming that the evaporation flow rate qV is 0.004 g/min
from the data of Fig. 12, the relative standard uncertainty
caused by evaporation is given as
u(qV ) 0.004 3.8 × 10−8
= √ = . (21)
q 60 × 1000 × 3 × q q
Even at long calibration time (over one hour), the standard
uncertainty in evaporation is reduced to 0.0007% by using this
lift-up closure mechanism. If the tank is in the lowered position,
 R. Doihara et al. / Flow Measurement and Instrumentation 17 (2006) 141–152
Fig. 13. Effects of evaporation at low flow rate (20 kg/h).
Fig. 14. Youden plot between PD meter and Coriolis flowmeter at 1800 kg/h.
the standard uncertainty caused by evaporation becomes
0.022%.
5.6. Repeatability of the calibration system
It has been confirmed that the repeatability of the calibration
system is smaller than the deviation in the flow meter
repeatability, because the Youden plot does not indicate any
correlation between the K -factors, as shown in Fig. 14.
In the prototype, the total of 20 calibrations can be conducted
within about one hour, because a single calibration sequence
can be completed within 150 s, plus the collection time, by
using the full automatic system and operation program. The
new system has demonstrated higher throughput than that of
a conventional weighing tank system. Of course, at low flow
rates, a longer calibration time will be needed, but we might
be able to reduce the amount of liquid collected along with
151
the collection time at low flow rates by using the results
of uncertainty analysis—if the uncertainty from the weighing
scale is sufficiently small relative to the flowmeter repeatability
or other error sources.
6. Conclusions
A new diverter based on a rotating double wing with a multi-
nozzle system has been developed. The weighing tank system
has been improved by some devices.
Experiments using the prototype have shown that the design
has many advantages. It has been confirmed that the rotation
motion and multi-nozzle system raise no problems. Through
rotation of the two wings, the double method enables easy
implementation of the test under ISO4185. In the weighing tank
system, a single action of the platform can prevent the escape
of vapor, ensure connection of the air tube, provide protection
against impact, and provide a system that is fail-safe against
overflow.
The new weighing tank system can calibrate over a wide
flow range (over 120:1), larger than that of a conventional
system. By using the double-wing method, the diverter timing
error is reduced at any flow rate and for any nozzle condition.
We are able to use a large flow rate with a small tank, which
makes the collection time short, because the uncertainty from
the diverter is certain to fall within a negligibly small value. In
addition, this design avoids evaporation and dead volume error
for calibrations conducted at a low flow rate over a long time.
It is expected that operation throughput will be relatively
high compared with previous systems, because the minimum
collection time can be reduced to a certain limit as long as
the flowmeter maintains repeatability. This calibration system is
well suited to fully automatic operation. Thus, in this prototype
it takes only 150 s, plus the collection time, to carry out one
calibration.
In summary, the liquid flow calibration facility using the
present system achieves accurate measurement, easy operation,
a wide flow range, high throughput, low cost, a small weighing
tank, easy maintenance, and safety.
References
[1] Pursley WC. The calibration of flowmeters. In: Flowmeasurement.
Measurement and Control 1986;(June):37–45 [special issue].
[2] Gowda V, Yeh TT, Espina PI, Yende NP. The new NIST water flow
calibration facility. In: Proc of 11th FLOMEKO. 2003 [Paper-140].
[3] Lederer T, Mathies N, Rose J, Stuck D. New test facility for large water
flowrates up to 1000 m3/h in a temperature range between 3C and 90C at
PTB — Berlin. In: Proc of 11th FLOMEKO. 2003 [Paper-108].
[4] Engel R, Baade HJ. New-design dual-balance gravimetric reference
system with PTB’s new ‘Hydrodynamic test field’. In: Proc of 11th
FLOMEKO. 2003 [Paper-122].
[5] Terao Y, Takamoto M. Uncertainty analysis of large water flow calibration
facility. In: Proceedings of the 4th international symposium on fluid flow
measurement. 1999 [Paper 15 02].
[6] Shimada T, Doihara R, Terao Y, Takamoto M, Stolt K, Andersson A. New
Primary standard for hydrocarbon flowmeters at NMIJ — International
comparison between NMIJ and SP. In: Proc of 11th FLOMEKO. 2003
[Paper-111].
[7] Buttle RS, Kimpton A. ESKOM’S flow calibration facility. In:
Proceedings of the 5th FLOMEKO. 1989. p. 1–10.
152 R. Doihara et al. / Flow Measurement and Instrumentation 17 (2006) 141–152
[8] The International Organization for Standardization. Measurement of
liquid flow in closed conduits — weighing methods. 1st ed. International
Standard. ISO 4185; 1980.
[9] Mattingly GE. Fluid mechanics measurements. 2nd ed. Washington (DC):
Taylor and Francis; 1996. p. 301–66.
[10] Baker RC. Flow measurement handbook. Cambridge University Press;
2000.
[11] Poeschel W, Engel R, Dopheide D, Kecke HJ, Praetor R, Weist N, et al.
A unique fluid diverter design for water flow calibration facilities. In:
Proceedings of the 10th FLOMEKO. 2000 [A9].
[12] Engel R, Klages U. A novel approach to improve diverter performance in
liquid flow calibration facilities. In: Proceedings of the 10th FLOMEKO.
2000 [A10].
[13] Shimada T, Oda S, Takamoto A, Nagai S. A novel diverter for liquid
flow calibration facilities. In: Proceedings of the 2001 NCSL international
workshop and symposium. 2001 [4B].
[14] Yeh TT, Yende NP, Johnson A, Espina PI. Error free liquid flow diverter
for calibration facilities. In: Proceedings of ASME fluids engineering
division summer meeting. 2002 [FEDSM2002-31085].
[15] Yeh TT, Yende NP, Espina PI. Theoretical self-error-cancelling diverters
for liquid flow calibration facilities. In: Proc of 11th FLOMEKO. 2003
[Paper-139].
[16] Shimada T, Doihara R, Terao Y, Takamoto M. Uncertainty analysis of
primary standard for hydrocarbon flow at NMIJ. In: Proceedings of the
12th FLOMEKO 2004. 2004.
[17] Shimada T, Oda S, Terao Y, Takamoto M. Development of a new
diverter system for liquid flow calibration facilities. Flow measurement
and Instrumentation 2003;14:89–96.
[18] The International Organization for Standardization. Guide to the
expression of uncertainty in measurement (GUM). International Standard;
1992.