Turbine Owmeter Response To Transitional Ow Regimes: Flow Measurement and Instrumentation November 2017
Turbine Owmeter Response To Transitional Ow Regimes: Flow Measurement and Instrumentation November 2017
Turbine Owmeter Response To Transitional Ow Regimes: Flow Measurement and Instrumentation November 2017
net/publication/321229975
CITATIONS READS
0 59
3 authors, including:
Benjamin Bizjan
University of Ljubljana
24 PUBLICATIONS 72 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Benjamin Bizjan on 02 March 2018.
Abstract
1 Introduction
Despite the constant development of novel flow measurement methods, the turbine-type
meters remain widely used in volume measurement of natural gas due to their accuracy and
reliability. A well calibrated and properly installed turbine flow meter is capable of
measurements with less that ±0.25% error (Tang [1]). Several publications have so far dealt
with turbine flowmeter calibration and measurement uncertainty analysis (Van der Grinten
[2], Ruiz et al. [3], Nerijus et al. [4]), as well as with numerical modeling of the flowmeter
operation (Osiadacz and Witos [5], Guo et al. [6]).
The majority of existing studies was focused on steady-state operation of the flowmeter
where the flow rate of gas is constant or changes at a very slow rate. While such a uniform
transport of natural gas is certainly desired, flow conditions in the pipelines are often far from
constant due to fluctuations in gas supply and demand, which are closely linked to
1
unpredictable environmental conditions (Geršak et al. [7]). The two main types of time
dependent flows are pulsating and intermittent flows (Cascetta and Rotondo [8]), but the
actual gas flow may be a combination of both, especially over longer periods of time. Pulsating
flows are typically generated by pulsating devices such as compressors and regulators and are
characterized by harmonic oscillations of the gas flow rate about its mean value, constantly
resulting in excessive volume readouts by gas meter, known as over-registration or
overestimation (Lee et al. [9]). The effect of flow pulsations on the measurement uncertainty
of turbine flowmeters was investigated and modeled by Cheesewright et al. [10] and Lee et
al. [9] as a function of pulsation frequency and amplitude, as well as flow meter size and flow
rate. As shown by Stoltenkamp et al. [11], acoustic perturbations (e.g. due to standing waves
in the system) may also cause substantial oscillations of the gas flow rate, leading to significant
measurement errors due to the poor dynamic response of turbine flowmeters.
The effect of pulsating flow disturbances can be partly reduced by a proper installation of the
flowmeter, including a sufficient length of straight pipe sections and the application of flow
settling devices such as tube bundles and flow conditioning plates (Mattingley and Leh [12],
Miller [13]). However, even a properly installed turbine flow meter will still be affected by
intermittent flow transitions, which cannot be sufficiently reduced in the system. These are
mostly induced by quick changes in the flow rate of gas, usually due to opening or closing of a
pipeline valve or sudden changes in gas demand. The rotational speed of a flowmeter turbine
typically follows the flow acceleration well, but exhibits a significant delay when slowing in a
decelerating flow (Cascetta & Rotondo [8], Tonkonogij et al. [14], Tonkonogij & Tonkonogovas
[15]) resulting in over-registration of gas volume. Despite the exponential reduction of the
turbine rotational speed [14] [15], the meter response duration may be relatively long
compared to the duration of the flow transition (especially if the flow is completely stopped),
leading to significant over-registration of gas volume. According to Cascetta and Rotondo [8],
a typical medium-sized turbine flowmeter may experience as much as 10000 on/off cycles or
other significant step changes in flow rate per year, consequently overestimating the gas
volume by more than 1%. In [14] and [15], response characteristics and corresponding
dynamic measurement errors of turbine gas meters are discussed in dimensionless form.
2
over-registered gas volume) will be presented as a function of flow meter size, flow rate and
duration of the intermittent step flow disturbance.
2 Experimental methodology
2.1 Experimental set-up
Experimental work presented in this paper was performed at the Laboratory for testing and
calibration of gas meters at the company Sarajevogas in Sarajevo, Bosnia and Herzegovina.
The laboratory is situated in the Institute SAGALAB officially designated by the National
Institute of Metrology of Bosnia and Herzegovina and has successfully participated on
EURAMET project 1296 of inter-laboratory calibration comparison in the range of flow rates
from 20 m3/h to 1000 m3/h, having approved calibration and measurement capabilities.
SAGALAB facility operates on the master meter principle where the meter under test (transfer
standard) is located downstream from the standard meters. Ambient air is sucked by a blower
and the flow rate is adjusted by regulation of the blower and electromotive valve. Testing
procedure is software controlled.
3
Fig. 1. Measurement setup for testing the turbine flowmeter dynamic response
Flow rate, i.e. rotational speed of the flowmeter turbine was measured by a high frequency
(HF) pulse signal p(t), for flowmeters of three different sizes (diameter D, Table 1) with
respective maximum flow rates Qmax (basically, the volume of gas flown through the meter is
proportional to the number of pulses). Additionally, the velocity of circulating air at the pipe
centerline was measured by a hotwire (HW) anemometer (Dantec MiniCTA 55P11 [16] as a
measure of the instantaneous air flow rate. The main advantage of HW anemometry is its very
good dynamic response to changing flow conditions, making it well suited for evaluation of
other methods' performance in transitional flows. Both the flowmeter signal and the HW
anemometer signal were sampled at 5000 Hz and acquired to a computer using National
Instruments measurement cards and LabVIEW software.
Table 1. Turbine flowmeters used in experiments. Ranges of Q and Re are provided for
unrestricted flow (fully opened valve)
D Qmax Q range Re range
# Type (m) (m3/h) (m3/h) (∙103)
1 G250 0.1 400 50-250 11.8-58.9
2 G1000 0.2 1600 300-1100 35.4-130
3 G2500 0.25 4000 1100-3000 104-283
For the purpose of comparison to the turbine rotational speed, HW-measured velocity was
also normalized as vn = v/vm. However, due to the fact that the HW signal exhibited significant
turbulence-induced oscillations, the signal v(t) was smoothed and down sampled to 100 Hz
sampling rate. Then, vm was computed as the time-averaged velocity after the flow rate
through the meter has been fully resumed following the valve reopening. Finally, both ωn and
vn were plotted as a function of time (a sample diagram is shown in Fig. 2). Note that since vn
was computed from measurements conducted at the pipe centerline, it is linearly proportional
to the maximum flow velocity of a velocity profile, but not to the mean flow velocity v Q / A
(Q is the gas volumetric flow rate and A is the pipe cross-section area) as the profile shape
changes with Reynolds number (Eq. (1)). In Eq. (1), νa is the kinematic viscosity of air.
vD
Re (1)
a
However, flows in all operating points were turbulent with Re = 1.2∙104…2.8∙105 (Table 1),
suggesting similarly shaped velocity profiles roughly following the 1/7th power law curve (Chen
[17]). Following the methodology presented in [18], we estimate the ratio vm / v to vary by no
more than 6% between operating points. For these reasons, no corrections were made to the
value of vn, assuming vn Q and vn Re .
For characterization of dynamic response of the turbine flowmeter to the step flow transition,
multiple regression measures will be introduced. Due to the lag of the turbine rotational
speed, the flowmeter over-registers the volume of gas between t0 and t1 and under-registers
it between t1 and t2. Over-registered volume V1 and under-registered volume V2 can be
calculated by integration – Eqs. (2) and (3). The constant kt = 1s introduced to allow for
dimensionless formulation of V1 and V2, which is mathematically more correct considering the
fact that ωn and vn are already dimensionless. Note that the integration of normalized
quantities ωn and vn yields a result with [s] unit. For the example presented in Fig. 2, the
valuesV1 = 21.6 and V2 = 0.93 were obtained, meaning that between t0 and t2, the gas volume
is over-registered by V1-V2 = 20.7, or an equivalent of dimensional gas volume transferred
through the flowmeter at fully opened valve in 20.7 seconds (5.18 m3). Since the interval
length t0 and t1 in which the valve was closed was not constant in all operating points, the
volume V50 (Eq. (4)) will also be introduced as the over-registered gas volume in the first 50
seconds after t0.
t1
1
V1
kt (t ) v (t ) dt
t0
n n (2)
t2
1
V2
kt v (t ) (t ) dt
t1
n n (3)
t0 50 s
1
V50
kt (t ) v (t ) dt
t0
n n (4)
A measure closely related to over- and under-registered gas volumes is the ratio θ between
the acceleration and deceleration time of the flowmeter turbine – Eq. (5). Typically, the value
of θ is well below 1, as the turbine accelerates quickly and decelerates slowly upon a step flow
transition.
6
t2 t1
(5)
t1 t0
Another characteristic parameter will be introduced as the halving time, τ1/2, required for ωn
to drop to 50% of its initial value at t0. As such, the time constant τ1/2 is an important measure
of the flowmeter mass inertia and mechanical losses. A dimensionless formulation of the
halving time will be designated as τ = τ1/2/kT. Note that in case that the decay of ωn is
exponential, τ1/2 is linearly proportional to the time constant of the 63.2% exponential decay
τe: τe = τ1/2/ln(2) ≈ 1.44 τ1/2.
From Figs. 3-5 it can be concluded that the curves of ωn resemble the shape of exponential
decay function (consistent with the findings of Tonkonogij & Tonkonogovas [15]). However,
8
the dynamic response of the normalized rotating speed ωn to otherwise very quick flow
deceleration is slow, with halving times τ1/2 between 6 s and 35 s and generally dropping with
Q (Fig. 6). Consequently, the transferred volume of gas is largely over-registered between t0
and t1. The volume V50 (Fig. 6), which marks the over-registered volume between t0 and t0 +
50 s, can be observed to drop with Q and increase with the flowmeter size. Also, in all
operating points, the under-registered volume V2 is much lower than the over-registered
volume V1 (V1/V2 = 12…60, consider Fig. 6). This leads to a conclusion that V2 is not a significant
issue in transitional flow measurement and even reduces the overall measurement error (V1-
V2) through partly offsetting V1.
i a0 Re a1 a2 (6)
9
In Eq. (6), Πi is the modeled variable (V50, θ or τ). Regression coefficients a0, a1 and a2 are
obtained by least-squares fitting of the power-law model to the measurement data. The
following models are obtained:
All models in Eqs. (7)-(9) have a high coefficient of determination (R2): 0.923, 0.931 and 0.960,
respectively. This indicates a good agreement between multiple regression models and used
measurement data. As demonstrated by Figs. (7)-(9) and corresponding Eqs. (7)-(9), all
modeled variables (V50, θ and τ) increase with the flowmeter size δ. This is consistent with the
findings of Cheesewright et al. [10] and Tonkonogij & Tonkonogovas [15] who reported longer
response times and larger over-registered volumes for larger turbine flowmeters. Such
behavior can be explained by the fact that larger flowmeters possess a higher ratio of inertial
mass forces to dissipative forces of aerodynamic drag and friction in bearings and other
mechanical elements [5].
Fig. 7. Dimensionless volume V50 as measured in experiments (data points) and modeled by
Eq. (7) – trend lines (R2 = 0.923)
10
Fig. 8. Acceleration/deceleration time ratio θ as measured in experiments (data points) and
modeled by Eq. (8) – trend lines (R2 = 0.931)
Speaking of the Reynolds number effect, it is evident that an increase of Re (or equally, the
ratio Q/QMAX for a flowmeter of a given size) causes a more rapid response (i.e. lower halving
11
time τ1/2 and acceleration time t2-t1) of the flowmeter turbine rotation to the flow transition
phenomenon, and consequently also lower measurement error due to gas volume over-
registration. At the same time, reduction in θ implies that the dynamic response of the
flowmeter becomes even more asymmetrical, with turbine acceleration up to 12 times faster
than deceleration. Faster acceleration with rising Re and Q/QMAX values can be explained by
the fact that a flow with higher velocity (when valve is reopened) exhibits a greater kinetic
energy available for acceleration of the flowmeter turbine, whilst at the same time the relative
effect of mechanical dissipative forces is reduced (Osiadacz [5]).
As previously elaborated, the flow transition induced in presented experiments is very fast,
which is often not the case in natural gas pipelines where the flow rate is adjusted more
gradually. Also, flowmeter response times in natural gas systems are lower due to significantly
higher operating pressures (and consequently Reynolds numbers). As a result, gas volume
measurement errors due to unsteady gas flow are not as extreme as in given experiments,
though still significant if large amplitude flow fluctuations are present in the system.
Nevertheless, very high R2 values of multiple regression models for V50, θ and τ suggest that
these models could be extrapolated beyond the experimental range of Reynolds numbers and
flowmeter sizes, thus potentially serving as a tool for assessment of measurement uncertainty
in the metering of transient natural gas flows.
4 Conclusions
In this paper, the dynamic response of turbine flowmeters to a step flow transition has been
investigated from the perspective of dynamic errors. Experimental results confirm a highly
asymmetrical flowmeter response characteristic to flow changes, which is consistent with
results previously reported in literature. Flow acceleration is not problematic as it is quickly
followed by acceleration of the flowmeter turbine. However, during the regime of flow
deceleration, the response of the flowmeter is much slower, causing the transferred gas
volume to be largely over-registered. Dynamic response during deceleration was found to be
faster for smaller flowmeters operating at relatively high Reynolds numbers (i.e. near the
maximum allowable flow rate). The same applies for the measurement error due to gas
volume over-registration.
With that said, multiple regression models for dimensionless response times and over-
registered volumes demonstrate a rather straightforward and well-correlated dependence on
the flow Reynolds number and the turbine flowmeter size. To avoid excess overestimation of
transferred gas volume, the duration of pipeline valve closing procedure should be of the
order of magnitude of 10 seconds or more. Slower valve closing is required when a turbine
12
flowmeter is operating at partial loads or at lower system pressures. Further research of
dynamic response characteristics should include measurements at elevated pressures and
with different types of flow variability (e.g. intermittent in form of a ramp or pulsating sine
wave function) to expand and verify the validity of multiple regression models presented in
this paper.
Acknowledgements
The authors acknowledge the contribution of the Laboratory of Metrology and Quality at the
University of Ljubljana, Faculty of Electrical Engineering in performing this study.
References
[1] Tang (2014) Pressure, temperature, and other effects on turbine meter gas flow
measurement. American School of Gas Measurement Technology, Houston.
[2] Van der Grinten, J.G.M. (1990) Error curves of turbine gas meters. The Netherlands
Measurement Institute, Delft.
[3] Ruiz, V., Pereira, M.T., Taira, M.N. (2013) Turbine flowmeter and viscosity effects of liquid
hydrocarbons. In: 16th International flow measurement conference, 2013, Paris.
Proceedings… (FLOMEKO, 2013). 5 p.
[4] Nerijus, P., Gediminas, Z., Eugenijus, M. (2015) Influence of Gas and Liquid Viscosity on
Turbine and Positive Displacement Meters Calibration. In: 17th International Congress of
Metrology, 2015, Paris. DOI: 10.1051/metrology/20150001003.
[5] Osiadacz, A.J., Witos, M. (2010) The analysis of factors affecting the operating
characteristic of turbine gas meter. Archives of Mining Sciences, 55(1), 163-180.
[6] Guo, S., Sun, L., Zhang, T., Yang, W., Yang, Z. (2013) Analysis of viscosity effect on turbine
flow meter performance based on experiments and CFD simulations. Flow Measurement and
Instrumentation, 34, 42-52, DOI: 10.1016/j.flowmeasinst.2013.07.016.
13
View publication stats
[7] Geršak, G., Bojkovski, J., Batagelj, V., Hudoklin, D., Begeš, G., Pušnik, I., Drnovšek, J. (2008)
On-site environmental conditions - do they really affect the measuring accuracy?. Journal of
testing and evaluation, 36(6), 1-5, DOI: 10.1520/JTE101866.
[8] Cascetta, F., Rotondo, G. (2015) Effects of intermittent flows on turbine gas meters
accuracy. Measurement, 69, 280-286, DOI: 10.1016/j.measurement.2015.02.0080263-2241.
[9] Lee, B., Cheesewright, R., Clark, C. (2004) The dynamic response of small turbine
flowmeters in liquid flows. Flow Measurement and Instrumentation, 15, 239-248, DOI:
10.1016/j.flowmeasinst.2004.07.002.
[10] Cheesewright, R., Atkinson, K.N., Clark, C., Ter Horst, G.J.P, Mottram, R.C., Viljeer, J. (1996)
Field tests of correction procedures for turbine flowmeters in pulsatile flows. Flow
Measurement and Instrumentation, 7(1), 7-17.
[11] Stoltenkamp, P.W., Bergervoet, J.T.M., Willems, J.F.H., Van Uittert, F.M.R., Hirschberg, A.
(2008) Response of turbine flow meters to acoustic perturbations. Journal of sound and
vibration, 315, 258-278, DOI: 10.1016/j.jsv.2008.01.051.
[12] Mattingley, G.E., T.T., Yeh (1991) Effects of pipe elbows and tube bundle on selected types
of flow-meters, Flow Measurement and Instrumentation, 2, 4-13.
[13] Miller, R.W., “Flow Measurement Engineering Handbook”, McGraw-Hill Book Company,
1996.
[14] Tonkonogij, J., Pedišius, A., Stankevičius, A. (2008) The New Semi-Experimental Method
for Simulation of Turbine Flow Meters Rotation in the Transitional Flow. World Academy of
Science, Engineering and Technology, 40, 208-213.
[15] Tonkonogij, J., Tonkonogovas, A. (2013) Analysis of nonlinearity of the turbine gas meters
time constant during step response. Mechanika, 19(5), 526-530.
[16] Dantec Dynamics: MiniCTA and Multichannel CTA Systems. Available at:
https://www.dantecdynamics.com/docs/products-and-services/fluid-mechanics/cta/
PI313_MiniCTA_MultichannelCTA_Anemometer_System.pdf. Accessed on 2017-03-15.
[17] Chen, C.L. (1991) Unified Theory on Power Laws for Flow Resistance. Journal of
Hydraulic Engineering, 117(3), 371-389.
14