You are on page 1of 9

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/236116040

Metrological performances of Mass Flow Controllers for dynamic gas dilution

Article  in  Accreditation and Quality Assurance · March 2013


DOI: 10.1007/s00769-013-0974-y

CITATIONS READS

9 331

3 authors:

Alessia Demichelis Guido Sassi


INRIM Istituto Nazionale di Ricerca Metrologica Politecnico di Torino
20 PUBLICATIONS   54 CITATIONS    53 PUBLICATIONS   363 CITATIONS   

SEE PROFILE SEE PROFILE

Maria Paola Sassi


INRIM Istituto Nazionale di Ricerca Metrologica
68 PUBLICATIONS   331 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

PCM-based components for building and construction View project

Coating for clay bricks View project

All content following this page was uploaded by Alessia Demichelis on 19 May 2014.

The user has requested enhancement of the downloaded file.


Metrological performances of mass flow
controllers for dynamic gas dilution

A. Demichelis, G. Sassi & M. P. Sassi

Accreditation and Quality Assurance


Journal for Quality, Comparability and
Reliability in Chemical Measurement

ISSN 0949-1775
Volume 18
Number 3

Accred Qual Assur (2013) 18:181-186


DOI 10.1007/s00769-013-0974-y

1 23
Your article is protected by copyright and
all rights are held exclusively by Springer-
Verlag Berlin Heidelberg. This e-offprint is
for personal use only and shall not be self-
archived in electronic repositories. If you wish
to self-archive your article, please use the
accepted manuscript version for posting on
your own website. You may further deposit
the accepted manuscript version in any
repository, provided it is only made publicly
available 12 months after official publication
or later and provided acknowledgement is
given to the original source of publication
and a link is inserted to the published article
on Springer's website. The link must be
accompanied by the following text: "The final
publication is available at link.springer.com”.

1 23
Author's personal copy
Accred Qual Assur (2013) 18:181–186
DOI 10.1007/s00769-013-0974-y

GENERAL PAPER

Metrological performances of mass flow controllers for dynamic


gas dilution
A. Demichelis • G. Sassi • M. P. Sassi

Received: 7 June 2012 / Accepted: 13 March 2013 / Published online: 24 March 2013
Ó Springer-Verlag Berlin Heidelberg 2013

Abstract The method of dynamically generating refer- Introduction


ence mixtures of volatile organic compounds at trace levels
requires the uninterrupted blending of the component flows The method of dynamically generating reference mixtures
during the generation time. The target flow rate uncertainty of volatile organic compounds (VOCs) at trace levels
depends on the application and the range of the concen- requires the uninterrupted blending of the components, i.e.,
tration generated. Accurate air flow rates used for dilution the mass flows of VOCs and dilution air [1] during the
can be realized by mass flow controllers (MFCs). The dilution time that generally is shorter than 1 h. Different
short-term stability of MFCs has been measured in order to schemes of dilutors can be adopted, diluting the VOC flow
assess their usefulness in dynamic gas dilutors. As shown by one or two separate, yet interfaced stages. The dilutor
here, MFCs are appropriate for the dynamic preparation of scheme used in the INRIM laboratory is schematically
diluted gas mixtures provided they are calibrated prior to shown in Fig. 1: F1 and F2 are the two main dilution air
each use, because their calibration is not stable over longer flows and Fint is the interface flow between the stages.
time periods. The calibration results of the low pressure To make the Fint control system work properly, a pres-
drop MFC proved its applicability in a wide range of sure drop from the first stage to the second one is
pressure drop. The uncertainty of the dilution flow rate necessary. If diffusion tubes [2] are employed, the pressure
resulted mainly from the limited short-term stability of the in the first line (p1 in Fig. 1) must be the lowest possible
MFC at low dilution flow rate or at short generation time. and cannot exceed 10 kPa. Consequently, a low pressure
Consequently, MFCs can provide highly accurate dilution drop (2–3 kPa) flow control system must be used for the
air flow rates with a relative uncertainty of 0.2 %, for flow inter-stage flow Fint.
rates larger than 70 % of the MFC’s maximum flow rate The contribution of the air flow rates to the uncertainty
and generation times longer than 5 min. of the VOC concentration can be considered as insignifi-
cant or negligible if their significance index [3], i.e., the
Keywords Mass flow controllers  VOC  Dynamic ratio between the flow rate contribution and the maximal
dilutor  Reference gas mixture contribution among the influence parameters in the uncer-
tainty budget is lower than 10 or 1 %, respectively. This is
related to the pre-defined target uncertainty of the VOC
concentration, which in turn limits the acceptable uncer-
tainty of the airflow rates. As a significance index of 10 %
indicates an insignificant contribution to the uncertainty
A. Demichelis (&)  G. Sassi  M. P. Sassi budget [3, 4], the target uncertainty of the air flow rate can
Istituto Nazionale di Ricerca Metrologica (INRIM), Turin, Italy be calculated from the uncertainty budget of the generated
e-mail: a.demichelis@inrim.it VOC concentration. The resulting target standard uncer-
tainty for the air flow rate is 0.2 % if the temperature of the
G. Sassi
Dipartimento di Scienza Applicata e Tecnologia, dynamic source is controlled to 0.1 K, or 0.1 % when
Politecnico di Torino, Turin, Italy controlling the temperature to 0.01 K [3].

123
Author's personal copy
182 Accred Qual Assur (2013) 18:181–186

p1
VOC
source
p1 regulating
valve
F1 p1 1-stage diluted VOC mixture
Atmospheric pressure
Purified air
401325 Pa Fint
F2 2-stage diluted VOC mixture
Atmospheric pressure

Fig. 1 Dynamic dilutor scheme. F1 and F2 are the two main dilution air flows, Fint is the interface flow between the stages, p1 is the mixture
pressure in the 1st dilution stage

Two types of devices can provide reliable air flow rates 100 ml min-1 under Standard condition (model LOW-Dp-
in a scheme as reported in Fig. 1: (1) devices such as the FLOW, F-101D-100-AAD-33-V ? F-004-AC-LU-33-V,
mass flow controller (MFC) allowing to adjust the flow rate Bronkhorst HI-TEC, Ruurlo (NL)) has been tested for Fint.
[5–7] and (2) fixed flow elements, e.g., sonic nozzles [8, 9]. Hereafter, they are referred to as F25A, F25B, and F01,
The flow rate accuracy of the first type is generally limited respectively. According to Bronkhorst specifications, the
by its stability [10, 11]. Thus, an in-depth understanding of optimum gas inlet pressure range is 200–300 kPa above
the MFC calibration uncertainty and their short-term sta- ambient pressure for F25A and F25B, and 2–3 kPa above
bility is required to use them in dynamic dilutors. ambient pressure for F01. Cylinders of synthetic air (Air
In the CCQM-K26b comparison [12], dynamic gas Liquide, air Alphagaz 1) with a declared purity of
standards have been prepared by permeation employing 99.999 % have been used without gas filter. The MFCs are
different flow control systems in gas dilution. An expanded intended to be used in the INRIM dynamic dilutors for
flow rate uncertainty (k = 2) of 0.5 % has been calculated reference mixtures for environmental studies at which in-
for a sonic nozzle (FMI laboratory) and a standard uncer- house generated purified air with a measured dew point
tainty of 0.1 % for an MFC when being calibrated against a lower than 213.15 K and particles with diameters of less
Brooks Vol-U-Meter (NMi, Netherlands). For a high- than 0.01 lm only was applied.
accuracy and variable gas flow dilutor using MFCs, a Flow rate measurements have been performed with both
typical expanded uncertainty of 0.4 % has been reported in the INRIM bell prover (160 l nominal capacity,
[13]. 0.2–400 l min-1) [14] and the INRIM piston prover (3 l
This paper investigates into the performances of MFC: nominal capacity, and a nominal flow range of
the calibration uncertainty (with emphasis on very low 0.1 ml min-1 to 2 l min-1) [15]. The experimental cali-
pressure drops), the variation of the calibration bias over bration setup is reported in Fig. 2. The setup has been
time, and the quantification of the short-term stability. tested before and after each measurement for leakages to be
The scope of this work is to experimentally verify the lower than 10 ll min-1.
usefulness of MFCs in dynamic gas dilutors. The results F25A and F25B have been calibrated at 10, 50, and 90 %
will lead to a more robust value than commonly reported, of their maximum flow rate by the bell prover (denoted in
for the MFC flow rate stability. It can be determinant in the Fig. 2 by number 4) with a relative calibration standard
choice of the gas control system. Furthermore, the inves- uncertainty of 0.1 %. F01 has been calibrated at 20, 60, 80,
tigation into the range of lowest gas pressure in the first and 100 % of maximum flow rate by the piston prover
dilution line contributes to optimizing the dilutor geometry, (denoted in Fig. 2 by number 5) with a relative calibration
thus allowing to miniaturize the VOC generation system. standard uncertainty of 0.05 %. The calibrations have been
repeated over a period of 16 months since shipping to
evaluate the calibration bias over time. F01 has been cali-
Materials and methods brated at maximum flow rate at a pressure drop range from
80 to 4500 Pa. The pressure drop has been changed varying
Two thermal MFCs (model EL-FLOW SelectÒ, F201CV- the upstream MFC pressure (effected by valve 3 in Fig. 2)
20 K-AAD-33-V, Bronkhorst HI-TEC, Ruurlo (NL)) at a and keeping a constant downstream pressure (atmospheric
nominal flow rate of 25 l min-1 under Standard condition pressure) as the dilutor configuration requires (Fig. 1).
(273.15 K and 101325 Pa) have been tested for F1 and F2. For the quantification of the short-term stability, F25A
A mass flow controller at a nominal flow rate of and F01 flow rates have been measured over 5 h as average

123
Author's personal copy
Accred Qual Assur (2013) 18:181–186 183

Fig. 2 Experimental setup for


PIC FIC
MFC calibration. 1 pure air
cylinder, 2 pressure reducer set FIR
at 401325 Pa, 3 p1 regulating n.1
valve, n.1 connected for bell
prover measures, n.2 connected
for piston prover measures, 4 2 F25A – F25B n.2 4
bell prover, 5 piston prover, 6
vent to atmospheric pressure, 6
PIC pressure indication and PIC FIC
control, FIC flow rate indication 1 FIR
and control, FIR flow rate
indication and registration, DPI
differential pressure indication
3 p1 F01 5
6 ΔPI 6

values of 10 s each. Flow rates of 10, 50, and 90 % of the


maximum were studied with F25A, while for F01 just the
maximum flow rate was applied. The flow rate data were
grouped in time intervals to model observation times s
between 20 s and 100 min, thus dividing the total of 5 h
into 900 to 3 intervals containing 3 to 900 average flow rate
values, respectively. In each interval, the relative standard
deviation sr of flow rate values was evaluated. As a con-
sequence for s = 1 min, there will be 300 values of sr. For
each observation time s, the 98th percentile of the normal
distribution of the sr data was calculated, i.e., the mean
value of sr data plus 2 times relative standard deviation of sr
data. This value has been demonstrated to be a more robust
estimator than the observed maximum value of sr data [3]
for the estimation of the short-term stability.
Fig. 3 F25A and F25B relative calibration bias over time. F25B
points are circled by the gray ellipses. Dotted lines refer to the target
Results and discussion flow rate standard uncertainty (0.2 %)

Calibration bias over time


F25B (vertical error bars in Fig. 3). The uncertainty stated
Taking the calibration flow rate of a MFC at the shipping in the manufacture’s calibration certificate for F25A and
time, Ft0, as reference value, the flow rate Ft, measured F25B is 1 % at maximum flow [10].
according to the INRIM calibration procedure at a later In the 16-month period, a calibration bias of up to 2.5
time, t, can be used to calculate a relative calibration bias B was observed, for measurements at 10–90 % of the maxi-
  mum flow rate it exceeded the target flow rate uncertainty
Fi  Ft0
B¼ ð1Þ (0.2 %) already after a few months. The calibration bias
Ft0
remained below the target uncertainty over the test period
B has been measured repeatedly over a 16-month period. when a MFC was operated at 50 % of maximum flow rate;
In Fig. 3, B values for F25A and F25B at 10, 50, and however, this hold just partly for the other tested MFC
90 % of maximum flow rate are reported as a function of which deviated beyond the manufacturers stated uncer-
the time t. Shipping time has been set as t = 0. Each value tainty (1 %) after a year. Similar results have been obtained
is the average of 4 measurements, the measurement for F01. Since the calibration bias exceed the target flow
repeatability standard deviation is lower than 0.04 %, and rate uncertainty, a MFC must be calibrated before (and
the INRIM calibration uncertainty is 0.1 % for F25A and after) each mixture generation.

123
Author's personal copy
184 Accred Qual Assur (2013) 18:181–186

Fig. 5 Frequency of the experimental relative standard deviation sr of


Fig. 4 Pressure drop Dp effect on F01 calibration standard uncer- the MFC flow rates at different observation times s (F25A at 90 % of
tainty at the maximum flow rate. The vertical error bars indicate the maximum flow rate). The y-axis reports the frequency of the event
standard calibration uncertainty of 0.05 % of the INRIM standard normalized by the total number of events

Low pressure drop effects relative standard deviation data of flow rate sr at any
given observation time s. The distribution of the sr data
A working pressure drop Dp between 2 and 3 kPa is rec- is reported in Fig. 5 for the flow rate measurements of
ommended for the MFC F01 by the manufacturer. Since F25A at 90 % of its maximum flow rate, for different s.
the VOC source, i.e., the diffusion cell, is directly con- The shape of the distribution depends on the observation
nected to the first stage of the dilutor, it is opportune to time s. For s longer than 1.5 min, i.e., with more than 9
limit its pressure in order to reduce gas leakage problems data per interval, a symmetrical shape is observed and a
and inaccurate pressure correction for the diffusion rate. normal distribution can be assumed with sufficient
F01 has been calibrated at different values of Dp from 80 to accuracy.
4500 Pa. In Fig. 4, experimental results are shown: each In Fig. 6, the short-term stability V is reported in
point is calculated as the average of four measurements, the dependence of the observation time s, at different mass
mean value of the flow rate over the tested pressure drop flow rates (10, 50, and 90 % of maximum flow rate for
range is reported as a horizontal line, the vertical error bars F25A and at maximum flow rate for F01). The few data per
indicate the standard calibration uncertainty of 0.05 % of interval at observation times lower than 1 min might result
the INRIM standard. in a weak estimation of the short-term stability V. How-
Figure 4 outlines that the flow rate values vary within ever, the dynamic generation of a reference VOC mixture
the calibration standard uncertainty of 0.05 % for a pres- generally requires more than 10 min.
sure drop between 100 and 4500 Pa. The Fisher statistical V slightly decreases with s for all the MFCs for all flow
test on experimental data of Fig. 4 outside and inside the rates. Thus, long observation times improve mass flow rate
recommended range of 2000–3000 Pa has shown that the stability. This aids in the identification of the optimum
experimental results belong to the same population with a observation time for any generation of dynamic gas mix-
confidence greater than 97 %. tures, e.g., the sampling time for a canister or the residence
A deviation of 2.5 % from the average value of flow rate time for flowing in a spectroscopic cell.
over the pressure drop range from 100 to 4500 Pa was mea- The best flow rate stability V has been observed at
sured for the mass flow rate at a pressure drop of 80 Pa. It can maximum flow rate while it approaches 3 % at 10 % of
be summarized that above 100 Pa, even outside of the sug- maximum flow rate. The analysis of V data for F25A
gested working range (2–3 kPa), the performance variation of shown in Fig. 6 has revealed a logarithmic dependency on
F01 is lower than the measurement uncertainty. Conse- the observation time s and mass flow rate F. The best
quently, the pressure drop has no effect on the calibration fitting result is reported as Eq. 2.
standard uncertainty of F01 for a Dp higher than 100 Pa.
ln V ¼ ð0:11 F  0:199Þ ln s  3:894 F  3:262 ð2Þ
Short-term stability where V is the flow rate stability, s* is the numerical value
of the observation time when expressed in min and F is the
The MFC short-term stability, V, has been calculated as fraction of the actual flow rate under Standard conditions
the 98th percentile of the normal distribution of the relative to the maximum flow rate.

123
Author's personal copy
Accred Qual Assur (2013) 18:181–186 185

as their quadratic sum. As previously discussed, no pres-


sure effect on the flow rate calibration standard uncertainty
has been considered.
The calibration standard uncertainty has been derived
from the traceable MFC calibration. To eliminate the
contribution of the calibration bias, it has been suggested to
calibrate the instrument before each use. The calibration
standard uncertainty gives a constant contribution on the
whole F and s ranges, while the short-term stability gives a
variable contribution expressed by Eq. 2. In Fig. 7, the
mass flow rate combined standard uncertainty u(F) is
reported as a function of the observation time s and of the
mass flow rate F for F25A (the curved surface). Expressing
Fig. 6 MFC short-term stability versus the MFC observation time s the mass flow rate F by the percentage of its maximum
for F25A and F01 at different percentages of maximum flow rate provides a direct comparison among MFCs of other span
The nominal maximum flow rate of F01 is lower than but with the same control valve.
10 % of the F25A maximum flow rate. Nevertheless, F01 The evaluated uncertainty for the flow rate is strongly
has shown a better short-term stability, i.e., around 0.2 % dependent on F (percentage of maximum flow rate) and on
versus the 2 % of F25A at 10 % of maximum flow rate. the observation time at low flow rates. Figure 7 reflects the
This may be due to the valve geometry, to the inherent behavior of the short-term stability which represents the
characteristic curve, and to the different ranges of the grade dominant contribution to the flow rate combined standard
of opening at which the valves operate. uncertainty. The horizontal plateau refers to the target flow
The flow rate stability of F25A is better than the target rate standard uncertainty of 0.2 %. As a result, the com-
standard uncertainty for the flow rate of 0.1 % or 0.2 % bined uncertainty of the flow rate is lower than the target
only for high flow rates. The short-term stability of F25B standard uncertainty only for high flow rates, i.e., on the
has been considered in this work close to that of F25A, due left side of Fig. 7. It follows that the target dilution flow
to the same valve geometry. rate uncertainty can be guaranteed working with the MFC
at flow rates larger than 70 % of its maximum flow rate and
Flow rate combined uncertainty for observation times longer than 5 min.

The combined uncertainty of the flow rate has been cal-


culated as the combination of the calibration standard Conclusion
uncertainty and the short-term stability at different obser-
vation times. Following the ISO/IEC Guide 98 (generally In this paper, MFC performances have been experimentally
known as GUM) [16], the contributions can be combined analyzed to investigate their use in dynamic gas dilutors.

Fig. 7 Flow rate combined


standard uncertainty
u(F) (curved surface) versus the 2.0
MFC flow rate F and
observation time s. The
horizontal plateau is the target 1.5
flow rate standard uncertainty of
Flow rate
0.2 %. Data are reported for
F25A 1.0 standard
uncertainty,
u(F) (%)
0.5

0.0
20 20
40
40
60 60
Observation time, 80 80 Flow rate, F (%)
τ (min) (percentage of maximum flow)

123
Author's personal copy
186 Accred Qual Assur (2013) 18:181–186

The following MFC conditioning procedure was employed: References


instrument warm up time of at least 1 h with the set flow
rate flushing before measurements, using purified air (dew 1. Nelson OG (1992) Gas mixtures: preparation and control. CRC
Press, Boca Raton
point lower than 213.15 K and particle diameters of less 2. ISO 6145 (2003) Gas analysis: preparation of calibration gas
than 0.01 lm), laboratory temperature control better than mixtures—dynamic volumetric methods, Part 8: diffusion
1 °C, ambient atmospheric pressure, and 30–70 % relative 3. Sassi G, Demichelis A, Sassi M (2011) Uncertainty analysis of
humidity. the diffusion rate in the dynamic generation of VOC mixtures.
Meas Sci Technol 22:105104–105111. doi:10.1088/0957-0233/
Experimental evidence has been given to the fact that 22/10/105104
MFCs need to be re-calibrated before and after each use to 4. Sassi G, Demichelis A, Sassi M (2009) A dynamic trace VOC
guarantee the uncertainty requirement for the generation of generator useful for global climate change study, in proceeding of
reference VOC mixtures. This suggests to provide a IMEKO XIX world congress, Lisbon, pp. 2602–2605
5. Tison SA (1996) A critical evaluation of thermal mass flow
handheld flow rate transfer standard that can be used before meters. J Vac Sci Technol A 14:2582–2591. doi:10.1116/1.
each generation. Alternatively, gas control systems based 579985
on fixed flow elements can be employed due to the easier 6. Viswanathan M, Kandaswamy A, Sreekala SK, Sajina KV (2002)
calibration procedure. Development, modeling and certain investigations on thermal
mass flow meters. Flow Meas Instrum 12:353–360
No systematic pressure effect was found to act on the 7. de Matos MAA, da Ferreira SV (2010) Gas mass-flow meters:
provided standard uncertainty of mass flow rate calibration principles and applications. Flow Meas Instrum 21:143–149
for the low pressure drop MFCs, in an inlet pressure range 8. Hayakawa M, Ina Y, Yukoi Y, Takamoto M, Nakao S (2000)
from 100 to 4500 Pa. Consequently, it will be possible to Development of a transfer standard with sonic Venturi nozzles for
small mass flow rates of gases. Flow Meas Instrum 11:279–283
construct a dilutor with a pressure limit in the first dilution 9. Bignell N, Choi YM (2002) Thermal effects in small sonic
stage down to 100 Pa which does not perturb the VOC nozzles. Flow Meas Instrum 13:17–22
mass flow rate from the generator. This allows to minia- 10. Brewer PJ, Goody BA, Gillam T, Brown RJC, Milton MJT
turize the VOC generation system with a low designed (2010) High-accuracy stable gas flow dilution using an internally
calibrated network of critical flow orifices. Meas Sci Technol
pressure loss. 21(11):115902. doi:10.1088/0957-0233/21/11/115902
A method for a conservative evaluation of the MFC 11. ISO 6145 (2003) Gas analysis: preparation of calibration gas
short-term stability over time (1–100 min) has been pro- mixtures—dynamic volumetric methods, Part 7: thermal mass
posed. The MFC short-term stability was found to be flow controller
12. International Key Comparison CCQM-K26b and Pilot Study
strongly flow rate-dependent and time-dependent at lower CCQM-P50b: Comparison of primary standards of Sulphur
flow rates. As a consequence, the MFC flow rate accuracy Dioxide in Synthetic Air. (2007) Metrologia 44(1A):08007. doi:
is limited by the flow rate short-term stability at lower flow 10.1088/0026-1394/44/1A/08007
rates. It is pointed out that the longer the observation time, 13. Goody BA, Milton MJT (2002) High accuracy gas flow dilutor
using mass flow controllers with binary weighted flows. Meas Sci
the better the MFC stability, granting better concentration Technol 13:1138. doi:10.1088/0957-0233/13/7/323
stability when a generated gas mixture is sampled for a 14. Baumann H, Cignolo G, Clausen M, Goria R (2006) EUROMET
longer time in a canister or in a detector cell. PROJECT 852: a comparison between Italian and swiss gas flow
In conclusion, the target standard uncertainties of standard in the range 0.3–25 m3 h-1. Metrologia 45:07011. doi:
10.1088/0026-1394/45/1A/07011
0.1–0.2 % for flow rates can be guaranteed using MFC in 15. Alasia F, Capelli A, Goria R, La Piana G, Cignolo G (2005) A
dynamic dilutors provided the flow rate control systems are primary standard piston prover for measurement of very small gas
calibrated before each mixture generation with the pro- flows: an update. Sensor Rev 25(1):40–45
posed conditioning procedure, employing a flow rate of 16. ISO/IEC Guide 98-3:2008, Uncertainty of measurement—Part 3:
guide to the expression of uncertainty in measurement
more than 70 % of the maximum and a generation time (GUM:1995)
longer than 5 min. Otherwise, dynamic dilution systems
need to be based on more stable flow rate control systems,
e.g., fixed flow elements.

123

View publication stats

You might also like