1-s2.0-S0957582015001792-main
-
Upload
cristianatanasiu -
Category
Documents
-
view
212 -
download
0
Transcript of 1-s2.0-S0957582015001792-main
-
7/25/2019 1-s2.0-S0957582015001792-main
1/10
Process Safetyand Environmental Protection 9 9 ( 2 0 1 6 ) 110
Contents lists available at ScienceDirect
Process Safety and Environmental Protection
journal homepage: www.elsevier .com/ locate /psep
Effective utilization of tracer gas in characterization
of underground mineventilation networks
Guang Xua,, Edmund C. Jongb, Kray D. Luxbacherb,c, Harold M. McNaird
a Western Australian School of Mines, Curtin University, Kalgoorlie 6430, WA, Australiab Virginia Center for Coal and Energy Research, Blacksburg, VA 24060, United Statesc Virginia Tech, Department of Mining and Minerals Engineering, School of Engineering, Blacksburg,
VA 24060, United Statesd
Virginia Tech, Department of Chemistry, Blacksburg, VA 24060, United States
a r t i c l e i n f o
Article history:
Received 23 June 2015
Received in revised form 15
September 2015
Accepted 1 October 2015
Available online 22 October 2015
Keywords:
Tracer gas
Underground mine ventilation
Gas chromatography
Evacuated gas sampler
CFD modeling
Experimental optimization
a b s t r a c t
Tracer gases are an effective method for assessing mine ventilation systems, especially
when other techniques are impractical. Based on previously completed laboratory and field
experiments, thispaper discussessomecommon andchallenging issuesencountered when
using tracer gases in underground mines. The discussion includes tracer release methods,
sampling and analysis techniques. Additionally, the use of CFD to optimize the design of
tracer gas experiments is also presented. Finally, guidelines and recommendations are pro-
vided on the use of tracer gases in the characterization of underground mine ventilation
networks. This work has informed the practical use of tracer gases in mines, and this body
ofknowledge is expected to contribute to more efficient and more common use of tracer
gasesby mineengineers, which willallowforbetter characterizationof mineventilationsys-
tem andimproved safety. Thefindings canalso be used when using thetracer gastechnique
in the evaluation of atmospheric environment and air quality investigation in buildings.
2015 The Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
1. Introduction
Ventilation is a fundamental to the engineering of under-
ground mines because it has considerable effects on health
and safety. Measurements of airflow underground are usu-allycarriedoutusingtraditional instrumentationsuchas vane
anemometers,hot-wireanemometers, pitottubes, and smoke
tubes.However, these methods arenotpracticalunder certain
circumstances in an underground mine. Examples include
the measurement of recirculation and leakage, flow in inac-
cessible zones, and flow with very low velocities. Sometimes
traditional instrumentation fails to provide accurate results.
For example, in a study that investigated jet fan effectiveness
in dead headings, tracer gas results were found to be more
accurate than theresultsofsmoke tubes (Timko andThimons,
1982). Therefore, tracer gas techniques are a valuable tool for
Corresponding author. Tel.: +61 8 90886113.E-mail addresses: [email protected], [email protected](G. Xu).
accurately measuring airflow in situations where traditional
methods cannot be employed and providing information to
characterize underground mine ventilation.
The tracer gas technique is a useful and versatile tool
for studying mine ventilation systems with a long history ofapplication. The Bureau of Mines (Thimons and Kissell, 1974)
conducted a series of tracer gas tests using sulfur hexafluo-
ride SF6 and proved the usefulness of tracer gas techniques in
measuringrecirculation, air leakage, airflowin large cross sec-
tion, lowflow velocity, andtransit air time. Grenier et al. (1992)
used tracer gases to analyze the spread of dust in a fluorspar
millingplant. Theresults indicatedthat tracer gases behave in
the same physical manner as respirable dust and can be used
to find patterns in dust movement within a ventilation sys-
tem. Tracer gas has been accepted bythe mining industry as a
viableventilation survey tool. More examples that used tracer
http://dx.doi.org/10.1016/j.psep.2015.10.0010957-5820/ 2015 The Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
http://www.sciencedirect.com/science/journal/09575820http://www.elsevier.com/locate/psepmailto:[email protected]:[email protected]://localhost/var/www/apps/conversion/tmp/scratch_7/dx.doi.org/10.1016/j.psep.2015.10.001http://localhost/var/www/apps/conversion/tmp/scratch_7/dx.doi.org/10.1016/j.psep.2015.10.001mailto:[email protected]:[email protected]://crossmark.crossref.org/dialog/?doi=10.1016/j.psep.2015.10.001&domain=pdfhttp://www.elsevier.com/locate/psephttp://www.sciencedirect.com/science/journal/09575820 -
7/25/2019 1-s2.0-S0957582015001792-main
2/10
2 Process Safetyand Environmental Protection 9 9 ( 2 0 1 6 ) 110
gas to investigate various ventilation problems can be found
in Timko and Thimons paper (Timko and Thimons, 1982).
Compared to the uses of tracer techniques in under-
ground mine, these techniques are wildly used in the
atmosphere studies and environmental monitoring applica-
tions (Vanderborght and Kretzschmar, 1984). Pekney et al.
(2012)useda perfluorocarbontracer thatwasinjected together
with CO2 to monitor the leakage of the carbon sequestrationtechnology. Connan et al. (2011) evaluated two computation
models for the prediction of atmospheric dispersion by com-
paring them to a SF6 tracer study. Belalcazar et al. (2009, 2010)
used tracer gas and CFD modeling to study the road traffic
emission factors (EFs), and showed that it is possible to esti-
mate the EFs from tracer studies.Jayaraman et al. (2006) used
tracer gas in an indoor space experiment for the purpose of
evaluating the accuracy of different CFD models. As most of
these studies can also be related to the problems in the mine
environment, it is evident that the tracer gas technique has
great potential in the research of mine safety and health.
An ongoing research project that involves the selection of
novel tracer gases for mine ventilation, the development of
a methodology to use tracer gases and computational fluid
dynamics (CFD) modeling to analyze, predict, and confirm the
underground ventilation status together with the location of
the damage, and finally validate the developed methodology
in thelaboratoryandinthefield are currentlybeingconducted
at Virginia Tech. Details of this work have been published in
severalforums (Jong et al., 2012; Pattersonet al., 2010; Xu et al.,
2013, 2015).
The focus of this paper is to provide some useful findings
for the use of tracer gas based on authors experience and
practice. As this research progressed it was evident that there
are few resources in the literature that provide the practical
aspects ofconductingtracer gasstudies inmines.Someessen-
tial aspects of the tracer gas technique are discussed as well
as new findings and recommendations, and studies in the lit-
erature are referenced as well. Using CFD modeling to design
tracer gas experiments is also presented in this paper. Some
modeling examples are provided to illustrate how CFD can
help to determine the optimized tracer release and sampling
locations, the release rate and duration, and eventually help
to achieve desired results. Tracer gas experiments are time
and resource consuming in underground mines, the findings
and recommendations provided in this paper can be used by
other researchers and industries for the design of tracer gas
experiments more efficiently with less trial and error.
2. Tracer gas techniques
2.1. Choices oftracers
Sulfur hexafluoride (SF6) is a widely accepted standard tracer
gas that has been used in mine ventilation studies. It was
first used in the building ventilation system sixty years ago
(Dick, 1950) and has been widely used for ventilation analy-
sis both in buildings and underground mines (Kennedy et al.,
1987). SF6 is non-toxic (Lester and Greenberg, 1950), and the
Occupational Safety and Health Administrations (OSHA) Per-
missible Exposure Limit (PEL) andtheAmerican Conference of
Governmental IndustrialHygienistss (ACGIH)ThresholdLimit
Value (TLV) for SF6 is 1000ppm (OSHA, 1993). The amount
of SF6 released to a mine is generally much less than either
the PEL or TLV limit. SF6 can be detected accurately using
gas chromatography (GC) in concentrations as low as parts
per trillion. It is also odorless, colorless, chemically and ther-
mally stable, and not found in the natural environment. It is
not measurably adsorbed on sandstone and coal. These are
all desirable properties as a tracer gas (Thimons and Kissell,
1974). There are a wide range of applications of SF6 in under-
ground mines,including the study of ventilation recirculation,
fan effectiveness, and flow rates estimation (Thimons et al.,
1974;Thimons andKissell, 1974), themeasurementof airleak-
ageand dust control effectiveness (Timkoand Thimons, 1982),
themeasurement of turbulentdiffusion (Arpa et al., 2008), the
study of methane control (Mucho et al., 2000), and the appli-
cation of SF6 to characterize underground mine ventilation
systems after emergency (Xu et al., 2013, 2015). Some alter-
native gases have also been used as mine ventilation tracers,
such as nitrous oxide and helium. However, because they are
not easily detected, large amounts need to be released thus
causing transportation problems and difficulty in achieving
stable flows (Kennedy et al., 1987).
It has long been realized that multiple tracer gases can
add flexibility to ventilation surveys in many ways. Multiple
gases notonly allowthe release of different tracers atdifferent
points without increasing the number of collected gas sam-
ples, but also provide more information because the source
of each tracer in one sample can be identified and air flows
in different zones can be investigated simultaneously with
multiple tracers. Once a tracer is released to the mine, it may
take days to weeks for the tracer background to be reduced
to a level that will not affect the next test. However, if mul-
tiple tracers are available, a different tracer can be used to
conduct another test right after the previous test. Although
the advantages are apparent, the use of the multiple tracer
technique is still not common in underground mines. One
key requirement for identifying other tracer gases is that the
gas should be measurable by the same method being used
for SF6, which is the most commonly used tracer gas. SF6 is
commonly analyzed by GC, so it would be better if the other
tracers were able to be analyzed by the same GC method.
Using thesame GCmethod, theadditional tracers shouldhave
similar sensitivity to SF6 as well as be able to be separated
from SF6. Kennedy et al. (1987) investigatedsixFreons that are
promisingcandidatetracer gases.They found that only Freon-
13B1 (CBrF3) and Freon-12 (CF2Cl2) were within two orders of
magnitude of the sensitivity of SF6 when analyzed using a
GC with an electron capture detector. The other four Freon
gases were several orders of magnitude less sensitive. CBrF3and CF2Cl2were tested in the field and it was concluded that
they perform well as mine ventilation tracer gases and are
comparable toSF6. Battermanetal. (2006)usedhexafluoroben-
zene (HFB) and octafluorotoluene (OFT) for indoor ventilation
tracers. However, those two tracers were not simultaneously
analyzed with SF6. Have used propane (C3H8) as a tracer gas
to evaluate the ventilation system of a full face tunnel bor-
ing machine, but pointed out that it cannot be used in coal
mines (Stokesand Stewart, 1985). Patterson, (2011) researched
the selection of novel tracer gases that can be used in mines
together with SF6. Freon 14(CF4), C3F8, and PMCH(C7F14) were
tested on different columns using various GC methods. CF4and C3F8 were found to have much less sensitivity than SF6on the tested columns, with similar retention times to SF6.
PMCH was reportedto beanappropriatetracer ifused together
with SF6. A GC protocol was developed that can be used to
analyze SF6 and PMCH concurrently on an HP-AL/S capil-
lary column. One drawback of the protocol is the long 18min
-
7/25/2019 1-s2.0-S0957582015001792-main
3/10
Process Safety and Environmental Protection 9 9 ( 2 0 1 6 ) 110 3
Fig. 1 Different flow meters.
analysis time,which process canpotentiallybe optimizedand
shortened.
In summary, as for the choice of tracers in underground
mines, SF6 is no doubt the best choice. CBrF3, CF2Cl2, andPMCH (C7F14) can be used together with SF6. CBrF3and CF2Cl2have both been successfully used in a mine. PMCH is a novel
underground mine tracer gas. The authors research group is
developing method to apply it as additional tracer that can be
used together with SF6.
2.2. Tracer gas release technique
Accurate and precise release of tracer gas is critical to con-
ducting a rigorous study; there must be high confidence in
the mass or rate of gas released to the system in order to
achieve meaningful analysis of results. In the application of
mine ventilation, the release methods also need to be reason-
ably portable and easy to operate. There are two commonly
used tracer gas release techniques: pulse-injection, which is
based on the injection of a short duration of tracer gas, and
constant-injection, which is based on the continuous injec-
tion of tracer gas. Pulse-injection is easier and quicker to
deploy, butrequires a large numberof measurementsto define
the resultant curve. Continuous injection is advantageous in
allowing evaluation of airflow stability (Stokes and Kennedy,
1987).
Tracer gases can be released in a controlled manner using
various methods. For pulse injection, a known mass of tracer
gas in a balloon or syringe can be used and injected to the
mine. Or directly release tracer gas from a pressurized con-
tainer and determine the weight loss of the container after
release. For example,in a studyconductedby Widiatmojo etal.
(2015) using tracer gas to investigate theairdispersion charac-
teristic in underground mine ventilation, tracer gas (SF6) was
releasedin balloons, andthemass of tracer gaswas measured
by the weight difference after and before the balloons were
filled. An example of using a syringefor pulse injection can be
found (Stokes and Kennedy, 1987). However,it is hard for these
methods to release in a controlled rate. Flow meters and per-
meation tubes serve as more accurate controlled rate tracer
gas release methods, and they can be used for both pulse and
constant injection.
Soap bubble flow meter (Fig. 1A), rotameter (Fig. 1B), and
electrical flow meter (Fig. 1C) are commonly used flow meters
that can be used to measure and to control the flow rate.
The soap bubble flow meter measures the flow rate of tracer
release, but accessory instruments are needed to control the
flow rate of gas. For gases contained in a compressed gas
bottle, a two stage regulator is often attached to the bottle.
Capillary tubing may be used to connect the regulator and
the soap bubble flow meter adding more resistance to achievebetter gas flow control. The two stage regulator controls and
maintains the delivery pressure to the flow meter as long
as gas tank pressure is greater than the delivery pressure.
Selecting different capillary tubing and adjusting the delivery
pressure affects the flow rate. The described setup provides
flow stability unaffected by atmospheric pressure (Grenier
et al., 1992).
Variable area flow meters, also known as rotameters,
provide a different set of controls. The design is based on the
variable area principle and indicates as well as controls the
rate of flow. Positioned vertically, the gas flow lifts the float in
the flow tube. For a constant flow rate, the float will be at a
stationary position, which corresponds to a point on themea-surement scale that indicates the flow rate. The flow rate can
bechangedby adjustingthedelivery pressure ormanipulating
the valve on the flow meter. Although a specific meter can be
purchased for a certain gas, the air flow meters are the most
commonly found in the market. The air flow meter can be
used for other gases with a correction factor provided by the
manufacture.
An electrical mass flowcontroller canbe used to accurately
control the release quantityof a tracer. Using a mass flow con-
troller, the flow rate can be adjusted using the digital control
panel. It requires minimal adjustment to achieve the desired
flow rate compared to the other types of flow meters.
The permeation device is a commonly used tracer releasesource for volatile compounds. The basic concept is to seal
a certain amount of tracer in an impermeable tube with per-
meable material at one end (as shown in Fig. 2). The emission
ratecanbe determined byweighingthe prepared tube at inter-
vals of several days until equilibrium is reached. Theemission
rate is relativelystableif the temperature is constant (Mailahn
et al., 1989). Dietz and Cote (1982) described a perfluorocar-
bon (PFT) source in which a known mass of PFT is injected
Fig. 2 Schematic of a permeation device.
-
7/25/2019 1-s2.0-S0957582015001792-main
4/10
4 Process Safetyand Environmental Protection 9 9 ( 2 0 1 6 ) 110
into a fluoroelastomer plug and crimped in a metal shell. PFT
will diffuse from the end of the plug at a known rate that is
inversely proportional to the square root of time for the emis-
sion of the first 5060% of the original amount of PFT.Johnson
et al. (2007) described the preparation of a permeation device
they used for continuous and constant release of SF6. It was
constructed from brass rod, Teflon, a frit, and a swaglok nut.
Although their design was not used in a mine tracer test, the
method can be used for the release of SF6 and other tracer
gases in mines. Batterman et al. (2006) described an updated
methodfora constant injectiontechniquewith miniaturePFT
sources. The method was shown to be reliable for measuring
indoor air exchange rates. The source was a diffusion con-
trolled release of saturated vapor in the headspace of a PFT
liquidcontainer.A diffusiontube was insertedinto theseptum
that sealed a glass vial partially filled with PFT. The emis-
sion rate was evaluated using a Fickian diffusion model and
tested by experimentation. The described sources mentioned
above can be easily modified for a wide range of applications.
For possibly application of the permeation device to mine
ventilationcharacterization,Jong et al., (2015b)designed a per-
meation plug release vessel (PPRV) for the release of PMCH.
The PPRV was made of an aluminum cylinder and plugged
with and oversized VMQsilicone plug. The release rate was
found to be constant, which was governed by a function of
temperature and plug thickness. The predicted release rate
was proved to be accurate though normal variations of baro-
metric pressure and insignificantly impacted by the PMCH fill
level. This design was later tested in both a laboratory con-
trolled turbulent environment (Jong and Luxbacher, 2015) and
an underground longwall mine (Jong et al., 2015a), in which it
was proved to be applicable to underground ventilation appli-
cations including long-termleakage and gob flow monitoring.
Thepermeationdevice hasthe advantagesof rapidsetup, ease
ofuse, anddeployment inpreviously restrictedareas, andthus
has great application potentials.
2.3. Tracer gas sampling methods
There are two categories of gas sampling methods: collecting
samples for laboratory analysis and collecting samples for
immediate analysis (Peach and Carr, 1986). The first category
is commonly used in underground tracer gas studies, which
includes gas sampling bags, hypodermic syringes, and evacu-
ated containers.
Gas sampling bags have been successfully used for a
number of years to collect gas samples and make gas stan-
dards. Gas sampling bags are available commercially and
come in a variety of sizes and shapes, and are made from
a number of materials, such as PVC (polyvinyl chloride) or
Tedlar (polyvinyl fluoride). One needs to pay special attention
to the material of the sampling bag because the sampled
gas may be reactive, adsorptive, absorptive, or diffusive with
the bag materials (Peach and Carr, 1986). There are many
applications of using gas sampling bags for collecting gas and
vapor samples (Schuette, 1967). It usually requires a pump
to inject gas into the bag, so it is not a very convenient tool
for dynamic gas sampling. Kennedy et al. (1987) tested the
tightness of TEDLAR gas sampling bags and no detectable
degradation of gas samples was found in a 24h period. The
storage time is typicallyno more than 24h, so analysis should
be conducted as soon as possible,unless storage experiments
indicate a longer storage time.
Hypodermic syringes satisfy most underground sampling
requirements. Syringes are inexpensive, easy to carry, and
hard to break.However, their short storage time restricts their
use. For example, in one study, gases such as carbon diox-
ide are lost rapidly due to permeation (Freedman et al., 1975).
It is standard practice to fill and evacuate the syringe twice
before drawing so that the gas left in the syringe does not
contaminate the gas sample. The syringe can be sealed witha vinyl cap (Grenier et al., 1992). The tightness of hypoder-
mic syringes have been tested by Kennedy et al. (1987). In the
study, six syringes were filled with certain concentrations of
SF6, CBrF3, and CF2Cl2, and left for 24h. The concentrations
did show a significant reduction, with a loss of 0.51% for SF6,
2.53%forCBrF3, and 6% for CF2Cl2. These syringes wereemp-
tied and refilled with the same standard gas, left for another
24h, and tested again for the tracer gas content. The loss of
tracer was reduced to half of the original percentage after the
first filling. This indicated that the loss was not only due to
permeation through the walls or leaks in the seals but also
due to adsorption of the tracer onto the syringes. Like samp-
ling bags, samples generally should be analyzed on the same
day they are taken.
Vacutainers have long been used for sampling mine gases.
A Vacutainer is an evacuated glass or plastic tube-shaped ves-
sel and is capped with a self-sealing rubber septum. Such
containers are commonly used to take blood samples. The
advantages of Vacutainers are that they are small, light-
weight, economical, convenient, and simple to use (Peach
and Carr, 1986). The Bureau of Mines officially adopted the
Vacutainers fortaking mine gassamplesbecause they arecon-
venient and can obtain consistent results (Freedman et al.,
1975). Freedman et al. (1975) tested the 10ml Vacutainers for
usein mines. In their study, a device was described which can
evacuate up to 56 Vacutainers to a few millimeters of pres-
sure. It was found that for stored gas samples, CO2 shows
substantial loss in concentration over the 41 days due to per-
meation. TheCO concentration levelgradually increased up to
50ppm over time in factory supplied or completely evacuated
Vacutainers. This phenomenon is unexplained. They recom-
mended that the storage of evacuated Vacutainers should not
exceed 1 to 2 months if low levels of CO can cause interfer-
ence. If precise CO2 level is of interest in collected samples,
the analysis should be done within 1 week after samples are
taken.
The 10ml Vacutainer was also chosen as the sampling
method for our experiments. These evacuated containers are
not completely evacuated at the time purchased and a small
amountof gas (air) is still present. However, thecontainersare
evacuated to a fixed anddesignated pressure (Bush, 2009). The
factory evacuation is designed to draw 10ml of blood. Prelim-
inary tests indicate they draw 9.5 ml of water. The left three
Vacutainers in Fig. 3 illustrate this result. It was noted that
the capability of drawing water to be reduce slowly with time.
The actual Vacutainer capacity measured by water displace-
ment is 12.9ml. Therefore, the Vacutainers were re-evacuated
in the laboratory to improve the sampling accuracy. The right
threeVacutainers in Fig. 3 show theresultof laboratory evacu-
ation, whichimprovedthecapabilityof Vacutainersof drawing
12.5 ml of water. This means 0.4 ml of air is left in each Vacu-
tainer, which is 3.1% of the actual capacity. This will make
the measured gas concentration 3.1% less than the actual gas
sample concentration, but it is acceptable for most tracer gas
analysis purposes.
-
7/25/2019 1-s2.0-S0957582015001792-main
5/10
Process Safety and Environmental Protection 9 9 ( 2 0 1 6 ) 110 5
Fig. 3 Vacutainers factory evacuation and laboratory
evacuation.
Vacuum Pump
Needle
Vacuum Gauge
Ball ValveVacuum Tube
PVC Pipe
Fig. 4 Evacuation apparatus schematic.
The evacuation system used in the laboratory is shown in
Fig. 4. Basically it is a vacuum pump connected to several nee-
dles so that several Vacutainers can be evacuated at a time.
Vacutainers are inerted onto each of the needles through theseptum for about 30s, and need to be pulled off very slowly,
which allows enough time for the septum to re-seal for a
high quality evacuation. Each needle connected to the sys-
temneeds to be replaced after about 5 Vacutainerevacuations
since it will dull and hard to penetrate the Vacutainer septa.
The re-evacuated Vacutainer shelf life was test by water
displacement over a period of 8 weeks. Several Vacutainers
were re-evacuated on day 0 using the above mentioned appa-
ratus. These Vacutainers were used later on different days to
test the shelf life. Its capacity change with time is shown in
Fig. 5. Each dot on this figure is an averaged value of three
Vacutainersthat wasre-evacuatedonday0.Thefigureshowed
that the capacity does not have obvious change in the first 5weeks, but dropped more than two percent from week 6 to 8.
This result suggests that, for a better sampling precision, the
Vacutainers should not be prepared (re-evacuated) more than
one month before taking samples.
95%
97%
99%
101%
103%
105%
0 7 14 21 28 35 42 49 56 63
Percentage
Time (Days)
Fig. 5 Re-evacuated Vacutainer capacity changewith time.
93.8%
96.9%96.6%
97.9%
96.8%
96.5%96.9% 97.1%
90.0%
91.0%
92.0%
93.0%
94.0%
95.0%
96.0%
97.0%
98.0%
99.0%
100.0%
0 1 2 3 4 5 6 7 8
PercentageoftheA
ctualSampleConcentration
Sample Time (s)
Concentration in Vacutainer
Expected Concentration
Fig. 6 Vacutainer sample time.
Although theVacutainer storage time is promising, an offi-
cial time has not been reported in the literature. An attempt
to test its storage time for SF6was conducted during a month
period and no obvious SF6concentration changes were found.However, the method used introduced significant amounts of
systematic errors that are difficult to control and quantify.
The Vacutainer sample time was also studied to determine
how fast the gas sample will fill the Vacutainers. The sample
time is defined as beginningwhen the rubber septum is punc-
tured bya needleendingwhenthe needle is removed fromthe
septum. This is important, especially for dynamic sampling,
because it determines the minimum sampling interval. As in
the previous test, for each test, three samples were taken and
the average concentration was used to compare with other
results. The test results are shown in Fig. 6. Given adequate
time for the Vacutainer to draw sample, the expected concen-
tration in the Vacutainer should be about 96.9% of the actualconcentration, as mentioned before. The 96.9% concentration
level is marked in the dashed line in Fig. 6. As can be seen, the
concentration in the Vacutainer reached the expected level
with sampling times longer than 2 s. Variations exist which
are likely due to error during Vacutainer evacuation and GC
manualinjection.Therefore,for the10 mlVacutainer, the least
sample time is recommended to be 2 s. Dynamic SF6samples
were successfully taken every 5 s during 8min period in our
underground tracer tests. In each 5 s, Vacutainer takes sample
for 3s and 2s are needed for changing to a new Vacutainer.
This is the fast minimum sampling interval we can achieve if
only one people takes samples manually.
2.4. Analysis oftracer gas
There are two main techniques for determining gas concen-
trations: infrared spectroscopy (IR) and gas chromatography
(GC). This paper focuses on the GCmethod which is one of the
most widely used analytical techniques for gas samples due
to its selectivity and sensitivity. A wide range of compounds
can be analyzed through the proper selection of columns and
detectors.
Gas chromatographs need to be calibrated over its
operating range to quantitatively measure a certain gas
concentration. Thus, tracer standards are needed for the cal-
ibration. The standards can be prepared in the laboratory or
purchasedas a certified mixture.To make thestandards in the
laboratory, a gas-tight syringe can be used to inject a small,
known quantity of tracer into a known volume of air or ultra-
pure nitrogen in a sealed container. An example of one of
-
7/25/2019 1-s2.0-S0957582015001792-main
6/10
6 Process Safetyand Environmental Protection 9 9 ( 2 0 1 6 ) 110
Fig. 7 Glass bulb for tracer gas standard preparation.
these containers is shown in Fig.7. With both valves open, thiscontainer can be flushed using ultra-pure nitrogen and filled
with it after both valves closed. To prepare standards for trace
concentrations, a serial dilution technique can be used. Serial
dilutions are completed by withdrawing a measured quantity
of the mixture prepared in the first step and injecting it into
another container. The procedure can be repeated until the
desired concentration is achieved. This method is time con-
suming and errors can be introduced at each stage of dilution.
Although this method can be used to produceuseable calibra-
tion curves, thepurchased certified gas standards can provide
better accuracy and repeatability (Leonard et al., 1984).
Split injection is a widely used injection technique for GC
analysis. It automatically reduces the sample size to preventthecolumnfrom overloadingbyallowingonly a fraction of the
sample enters the column. This fraction is defined as a split
ratio.A split ratio of 20:1 indicates that onepart of the sample
entersthe columnand20 partsexit theGC systemthroughthe
split vent (McNair and Miller, 1998). Theoretically, this means
1/21 of the total sample is analyzed. Again, theoretically, this
makes it possible to compare two results with the same GC
method but with different split ratio. For example, a sample
analyzed with a split ratio of 50:1 could be corrected to 20:1 by
multiplying a factor of 51/21. However, this split ratio calcula-
tion was found to be inaccurate due to the mechanical nature
of the splitting control. Fig. 8 shows the measured results of
the sameSF6sample with different split ratiobutallcorrectedtothe split ratio of20:1. Ascan beseen, thiscorrection method
fails to provide accurate results. Although the accuracy of the
split on different GC machines may vary, it is recommended
that a single split ratio is chosen and that corrections to the
data based on split ratio are avoided.
Since this correction methodis notalways reliable, thesplit
ratio should be held constant for all calibration standardsand
gas samples. A calibration curve is used to quantify the con-
centration in collected samples. To create a calibration curve,
theanalysis concentration range needs to be determined first,
and then start from a reasonable injection amount and split
ratio for the calibration. It is better to inject the highest con-
centration standard first to make sure the column will notbe overloaded. If it is overloaded, increase the split ratio or
reduce the amount of injected sample. Afterward, inject the
lowest concentration standard to ensure that the concentra-
tion is above the detection limit; otherwise, the split ratio
needs to be decreased or a larger amount of sample needs
to be injected. After the injection volume and split ratio are
determined, theseparameters need tobekept thesameforthe
rest of thestandards andtest samples that use this calibration
curve.
3. The use ofCFD model for tracer gasdesign
Tracer gas experiments are time consuming, especially if they
are conducted on a trial and error basis. Careful planning will
ensure meaningful data collected efficiently. CFD modeling is
a powerful tool that can be used for experimental design. This
tool cansubstantially reducetheeffort that is expended in the
field. In CFD modeling,tracers canbemodeled in severalways.
Thetwomost common methods are the two species transport
model or definition of a user defined scalar to model tracers.
Parameters can be easily changed in the CFD model, such asthe location and the release amount. This provides detailed
information to achieve the desired results, which can help
with the tracer gas experiment design. This section focuses
on the discussion of the determination of some parameters
using CFD modeling in the characterization of underground
mine ventilation networks.
One important parameter is that the expected concentra-
tion in collected samples needs to be within the detection
range of the GC. It is known that the concentration decreases
as the distance between the sampling and release location
increases. If these locations have been determined, the tracer
gas release rate and duration can be used to determine the
concentration level.The release rate is the most sensitive parameter that influ-
ences the concentration. Fig. 9 shows the SF6 concentration
profile results provided by one of the studies conducted by
theauthors. Thescenario modeled was a belt entry connected
with a crosscut entryshownin Fig. 12. Thereleaselocationis in
thecrosscut and is denoted as Release Point1.The sampling
location is 190 m away from the crosscut and is downstream
from thevelocityinlet in thebelt entry. SF6was releasedunder
different rates for 1 min in the CFD model. As can be seen in
Fig. 9, the basic shape of the concentration profile will not
change by changing the release rate, but the profile is taller
under larger release rate. The result can be used to find an
optimized release rate that can produce a tracer concentra-tion profile, which can be reasonably analyzed by GC or other
instruments.
154%
135%
100%
77%
63%53%
0%
20%
40%
60%
80%
100%
120%
140%
160%
SR 10:1 SR 15:1 SR 20:1 SR 30:1 SR 50:1 SR 100:1
PercentageoftheActualConcentration
Split Ratio
Measured Concentration Corrected to 20:1Split Ratio
Actual Concentration
Fig. 8 Same samplemeasured with different split ratio comparison.
-
7/25/2019 1-s2.0-S0957582015001792-main
7/10
Process Safety and Environmental Protection 9 9 ( 2 0 1 6 ) 110 7
0
50
100
150
200
250
300
0 50 100 150 200
SF6con
centration(ppm)
Time after tracer release
9 L/min
30 L/min
200 L/min
Fig. 9 Tracer concentration profile under different release
rate.
0
2
4
6
8
10
12
14
50 70 90 110 130 150 170 190 210 230 250
SF6Concentr
ation(ppm)
Time after release (s)
Release 10 s
Release 30 s
Release 1 min
Release 2 min
Fig. 10 SF6 concentration profile under different release
duration (Xu et al., 2015).
The release duration also influences theconcentration lev-
els in collected samples but is not as sensitive as the release
rate. Under certain circumstances, increasing the release
duration will not necessarily increase the concentration level.
In the same CFD model mentioned above, different tracer
release durations were modeled under the same release rate
(9L/min). As can be seen in Fig. 10, the 30s release increased
the concentration level when compared to the 10 s release.
However, increasing the release duration further, to 1 min
and 2min, did not increase the concentration level. That is
because the concentration reached its maximum under the
fixed release rate and the maximum concentration leveled
out. Although this is not always the case, the CFD results
can provide the data that can be used to pre-determine the
concentration level under certain release duration.
Another important parameter that CFD can help to deter-
mine is the shape of the expected concentration profile. The
farther the sampling location is from the release location, the
broader and flatter the profile becomes. If the release and
sampling locations are determined, longer release durations
can produce broader profiles as well. Fig. 10 shows that the
concentration profile becomes broader as the release dura-
tion increasedfrom 10s to2 min. Generally, a broader profileis
better, especially when the sampling method is not continu-
ous, such as syringeor Vacutainer sampling. A broader profile
allows enough samples to be taken to accurately resolve theprofile. If the profile is too narrow, for example it is only 10s,
but the sampling interval is 5s, only 3 samples maximum
can be taken within the 10s time. This makes it very hard
to accurately depict the concentration profile.
However, in some cases, too broad of a profile will not ben-
efit the tracer experiment not only because it takes longer for
sampling, but also because it weakens the profile characteri-
zation. As indicated from a CFD study presented in (Xu et al.,
2012), under the normal ventilation scenario, there are four
flow paths. If the tracer is released for 1 min, the profile is the
solid line displayed in Fig. 11, which has four peaks and each
of them represent one flowpath. However, if the release dura-
tion changes to 30min, the profile is the dashed line shown
in Fig. 11. It not only requires 30min more sampling time
to capture the entire profile, but the characterization is also
not as obvious as before. It is hard to tell that there are four
flow paths from this kind of tracer profile. In summary, too
short of a release duration may produce tracer profiles that
are hard to capture with certain sampling intervals; too long
of a release duration may weaken the tracer profiles charac-
terization. CFD modeling can be used to check if the release
duration is good enough for certain sampling intervals and to
characterize certain ventilation scenarios.
Changing the location of tracer release point will not only
change the profile peak arrival time but also change the pro-
file shape when the flow feature is complex. For example, in
the 2D CFD model mentioned before and shown in Fig. 12.
Two tracer release points were examined in the model: point
1 is 38m from the door, and point 2 is 2.5m from the door. As
can be seen in Fig. 13, due to the recirculating flow featured
in the zone at point 2, the downstream tracer profile not only
appears later and broader compared to release at point 1, but
the basic shape is also changed. The profile is much broader
and has a longer tail. In this case, the basic tracerprofile shape
is different when released in the same entry but at differ-
ent locations. The CFDmodel can identify these complex flow
zones,which provides information that helps to decide where
to release thetracer andgeta profile that is suitable for certain
purposes.
0
200
400
600
800
1000
1200
1400
1600
1800
0
50
100
150
200
250
0 10 20 30 40 50 60 70 80 90 100 110 120
SF6
Concentration(ppb)forRelease30
min
SF6C
oncentration(ppb)forRelease1min
Flow Time (min)
Release 1 min
Release 30 min
Fig. 11 Tracer profile under different release duration.
-
7/25/2019 1-s2.0-S0957582015001792-main
8/10
8 Process Safetyand Environmental Protection 9 9 ( 2 0 1 6 ) 110
Fig. 12 Tracer release at different locations (Xu et al., 2015).
0
2000
4000
6000
8000
10000
12000
14000
60 120 180 240 300 360 420
SF6Concentration(ppb)
Time (s)
Release 38 m from the door
Release 2.5 m from the door
Fig. 13 Tracer profile under different release location.
The examples shown above demonstrate that CFDis a very
powerful tool as an aid for tracer gas experimental design.
The modeling results canprovide informationfordetermining
some of the key experimental parameters mentioned above.
Using these optimized parameters to perform the actual on-
site experiment can help to avoid the trial and error process
and obtain the desired results efficiently. However, the CFD
model requires much more computational power compared
to other modeling methods, such as network modeling. Using
a high performance computer for the modeling is a solution
to reduce computational time. In addition, starting from a 2D
model can also save time tremendously. Although 2D flow
is totally different from 3D flow, and the 2D model cannot
provide as accurate a result as 3D, a 2D model can still provide
enough information to determine most of the influencingfac-
tors. After those factors have been determined, using a 3D
model to validate the 2D model can increase the confidence
andtheaccuracyof theresults.Finally, in cases where theflow
regime is notcomplex, andthesampling andcollection points
relative to the entry cross-section are not considered critical
to results, network modeling can even be used to understand
expected results. The advantage of network modeling is much
lower computational requirements along with less user skill.
4. Conclusions and discussion
The tracer gastechnique is a preciseand reliablemethodology
for characterizing underground mine ventilation networks.
However, it is time consuming and resource heavy, especially
when it is based on a trial and error basis. The guidelines
and recommendations provided in this paper are based on
the experiences and practices of the tracer gas research con-ducted over the last few years. These recommendations can
help other researchers and industries to reduce the effort in
conducting tracer gas experiments.
Several commontopicsin thetracer gastechniquesare dis-
cussed.As forthe choiceof tracers,CBrF3andCF2Cl2are found
to have been successfully used together with SF6 using the
same GC method. PMCH was also found to be a good tracer
for use with SF6. However, PMCH has not yet been proven in
underground mines.
As for the tracer release techniques, commonly used
flow meters are discussed and compared. Additionally, tracer
release by permeation tube is introduced although has yet to
be used in mines. It may prove to be a promising techniquefor tracer release in mines. These release methods control
the flow rate differently and can be used to serve specific
purposes.
There are several tracer gas sampling methods that have
been used in mines. Both gas sampling bags and hypodermic
syringes have a short storage time. Samples taken using these
methods should be analyzed within 24h. The material of the
sampling bags could affect its sampling capability for certain
gas samples. It was reported in the literature that the sample
loss inthehypodermicsyringesaredueto permeation through
the walls, leaks in the seals, and adsorption of the tracer onto
the syringes. Vacutainers have the longest storage time com-
pared to other sampling methods. However, substantial lossof CO2in stored samples was reported due to permeation. CO
concentration levels were foundto be graduallyincreasedover
time in evacuated Vacutainers in one instance. Storage time
and sample time for SF6 have not been officially reported in
the literature. An attempt to test its storage time for SF6 was
conducted during a month period and no obvious SF6 con-
centration changes were found. However, the method used
introduced significant amounts of systematic errors that are
difficult to control and quantify. The minimum sample time
of the Vacutainers is found to be 2s, however, the minimum
practical sampling interval for one person sampling manually
is 5s.
GC is a commonly used method for tracer analysis. The
preparation of gas standards in the laboratory was described,
which can be used to calibrate a GC when a certified gas mix-
ture isnotavailable. Thecorrection methodfrom oneinjection
split ratio to another is discussed and found to be inaccurate,
thus it shouldbe avoidedunlessexperimentsshowit isaccept-
ably accurate. A procedure was proposed for generation of a
calibration curve. These guidelines can be used to save time
and guarantee an accurate GC result.
CFD modeling can help with the tracer experiment design.
Examples are presented which illustrate how CFD modeling
helps to determine the important factors in a tracer exper-
iment, such as the release rate and duration, the expected
concentration profile, and the release location. After these
parameters areoptimized in theCFD model,the trial anderror
process canbe reduced andthe desired results canbe obtained
more efficiently. However, for a large scale 3D CFD model, the
computational time is considerable. Many of the parameters
-
7/25/2019 1-s2.0-S0957582015001792-main
9/10
Process Safety and Environmental Protection 9 9 ( 2 0 1 6 ) 110 9
can bestudied usinga 2Dmodel tosave time. Butthe2D model
cannot provide as accurate result as the 3D model because 2D
flow is totally different from 3D flow.
The aim of this work is to provide a practical guide for
people with technical expertise who want to apply tracer gas
techniques to mine ventilation. Although many impressive
applications of tracer gas techniques are available in the lit-
erature, few, if any sources are available that provide a guide
use of the method. This method can allow for characteriza-
tion of the ventilation system from an assessment of leakage
to the efficiency of designs for gas and dust dilution, and
provide engineers with additional tools for improving health
and safety. As tracer gas techniques are wildly used in the
atmosphere studies and environmental monitoring applica-
tions, this work can also be useful in these areas.
Acknowledgements
This publication was developed under Contract No. 200-2009-
31933, awarded by the National Institute for Occupational
Safety and Health (NIOSH). The findings and conclusions inthis report are those of the authors and do not reflect the offi-
cial policies of theDepartment of HealthandHuman Services,
nor does the mention of trade names, commercial practices,
or organizations imply endorsement by the U.S. Government.
References
Arpa, G., Widiatmojo, A., Widodo, N.P., Sasaki, K., 2008. Tracer gasmeasurement and simulation of turbulent diffusion in mineventilation airways. J. Coal Sci. Eng. (China) 14, 523529.
Batterman, S., Jia, C., Hatzivasilis, G., Godwin, C., 2006.Simultaneous measurement of ventilation using tracer gastechniques and VOC concentrations in homes, garages andvehicles. J. Environ. Monitor. 8, 249256.
Belalcazar, L.C., Clappier, A., Blond, N., Flassak, T., Eichhorn, J.,2010. An evaluation of the estimation of road traffic emissionfactors from tracer studies. Atmos. Environ. 44, 38143822.
Belalcazar, L.C., Fuhrer, O., Ho, M.D., Zarate, E., Clappier, A., 2009.Estimation of road traffic emission factors from a long termtracer study. Atmos. Environ. 43, 58305837.
Bush, V., 2009. The Evolution of Evacuated Blood Collection Tubes.Connan, O., Leroy, C., Derkx, F., Maro, D., Hbert, D., Roupsard, P.,
Rozet, M., 2011. Atmospheric dispersion of an elevated releasein a rural environment: comparison between field SF6 tracermeasurements and computations of Briggs and ADMSmodels. Atmos. Environ. 45, 71747183.
Dick, J.B., 1950. Measurement of Ventilation Using Tracer Gas.Heat. Piping Air Cond. 22 (5), 131137.
Dietz, R.N., Cote, E.A., 1982. Air infiltration measurements in ahome using a convenient perfluorocarbon tracer technique.Environ. Int. 8, 419433.
Freedman, R.W., Humphrey, W.G., Craft, R.L., 1975. Use ofVacutainers for Collection of Mine Atmosphere Samples.United States Bureau of Mines, Pittsburgh, PA.
Grenier, M.G., Hardcastle, S.G., Kunchur, G., Butler, K., 1992. Theuse of tracer gases to determine dust dispersion patterns andventilation parameters in a mineral processing plant. Am.Ind. Hyg. Assoc. J. 53, 387394.
Jayaraman, B., Finlayson, E.U., Sohn, M.D., Thatcher, T.L., Price,P.N., Wood, E.E., Sextro, R.G., Gadgil, A.J., 2006. Tracer gastransport under mixed convection conditions in anexperimental atrium: Comparison between experiments and
CFD predictions. Atmos. Environ. 40, 52365250.Johnson, K.A., Westberg, H.H., Michal, J.J., Cossalman, M.W., 2007.The SF6 tracer technique: methane measurement fromruminants. In: Makkar, H.P.S., Vercoe, P.E. (Eds.), MeasuringMethane Production from Ruminants. Springer, Dordrecht,The Netherlands, pp. 3367.
Jong, E.C., Luxbacher, K.D., 2015. An evaluation of aperfluoromethylcyclohexane (PMCH) permeation plug releasevessel (PPRV) in a controlled turbulent environment. Int. J.Min. Sci. Technol. 25, 243251.
Jong, E.C., Luxbacher, K.D., Karmis, M.E., Westman, E.C., 2015a.Field test of a perfluoromethylcyclohexane (PMCH)permeation plug release vessel (PPRV) in an undergroundlongwall mine. Min. Technol. (in press).
Jong, E.C., Peng, Y., Bradley, W.T., Luxbacher, K.D., Agioutantis, Z.,McNair, H.M., 2015b. Development of aperfluoromethylcyclohexane (PMCH) permeation plug releasevessel (PPRV) for tracer gas studies in underground mines.Process Saf. Environ. Protect. 95, 136145.
Jong, E.C., Underwood, S.W., Xu, G., Luxbacher, K.D., McNair, H.M.,2012. A technique for creating perfluorocarbon tracer (PFT)calibration curves for tracer gas studies. In: 14th U.S./NorthAmerican Mine Ventilation Symposium, Salt Lake City, UT.
Kennedy, D.J., Stokes, A.W., Klinowski, W.G., 1987. ResolvingComplex Mine Ventilation Problems with Multiple TracerGases. In: 3rd Mine Ventilation Symposium, University Park,PA, pp. 213218.
Leonard, J.J., Feddes, J.J.R., McQuitty, J.B., 1984. Measurement ofventilation rates using a tracer gas. Can. Agric. Eng. 26,
4951.Lester, D., Greenberg, L.A., 1950. The toxicity of sulfur
pentafluoride. Arch. Ind. Hyg. Occup. Med. 2, 348349.Mailahn, W., Seifert, B., Ullrich, D., Moriske, H.-j., 1989. The use of
a passive sampler for the simultaneous determination oflong-term ventilation rates and VOC concentrations. Environ.Int. 15, 537544.
McNair, H.M., Miller, J.M., 1998. Basic Gas Chromatography. JohnWiley & Sons, Hoboken, New Jersey.
Mucho, T.P., Diamond, W.P., Garcia, F., Byars, J.D., Cario, S.L., 2000.Implications of recent NIOSH tracer gas studies on bleederand gob gas ventilation design. In: SME Annual Meeting, SaltLake City, UT.
OSHA, 1993. Chemical Sampling Information.Patterson, R.R., 2011. Determination of a Novel Mine Tracer Gas
and Development of a Methodology for Sampling andAnalysis of Multiple Mine Tracer Gases for Characterization ofVentilation Systems. Department of Mining and MineralsEngineering, Virginia Tech, Blacksburg, VA.
Patterson, R.R., Bowling, J.R., Xu, G., 2010. Design of anexperimental apparatus for the testing of novel tracer gases inunderground mines. In: 13th United States/North AmericanMine Ventilation Symposium. SME, Sudbury, ON.
Peach, M.J., Carr, W.G., 1986. Air sampling and analysis for gasesand vapors. Occup. Respir. Dis., 4168.
Pekney, N., Wells, A., Rodney Diehl, J., McNeil, M., Lesko, N.,Armstrong, J., Ference, R., 2012. Atmospheric monitoring of aperfluorocarbon tracer at the 2009 ZERT Center experiment.Atmos. Environ. 47, 124132.
Schuette, F.J., 1967. Plastic bags for collection of gas samples.Atmos. Environ. 1, 515519.
Stokes, A.W., Kennedy, D.J., 1987. A real-time tracer gasanalyzeran investigational tool for mine ventilation studies.Min. Sci. Technol. 5, 187196.
Stokes, A.W., Stewart, D.B., 1985. Cutting head ventilation of a fullface tunnel boring machine. In: 21st Int. Conf. Safety in Mines.Mines Research Institutes, Sydney, Australia.
Thimons, E.D., Bielicki, R.J., Kissell, F.N., 1974. Using sulfurhexafluoride as a gaseous tracer to study ventilation systemsin mines. U.S. Bureau of Mines, Pittsburgh, PA.
Thimons, E.D., Kissell, F.N., 1974. Tracer gas as an aid in mineventilation analysis. Bur. Mines.
Timko, R.J., Thimons, E.D., 1982. Sulfur hexafluoride as a mineventilation research toolrecent field applications. U.S.
Bureau of Mines, Pittsburgh, PA.Vanderborght, B., Kretzschmar, J., 1984. A literature survey ontracer experiments for atmospheric dispersion modellingstudies. Atmos. Environ. 18, 23952403.
Widiatmojo, A., Sasaki, K., Sugai, Y., Suzuki, Y., Tanaka, H.,Uchida, K., Matsumoto, H., 2015. Assessment of air dispersion
http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0005http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0005http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0005http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0005http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0010http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0010http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0010http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0010http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0015http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0015http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0015http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0020http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0020http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0020http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0025http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0030http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0030http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0030http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0030http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0030http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0030http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0030http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0030http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref1135http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref1135http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref1135http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0035http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0035http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0035http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0035http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0040http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0040http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0040http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0045http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0045http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0045http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0045http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0045http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0045http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0050http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0050http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0050http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0050http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0050http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0055http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0055http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0055http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0055http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0055http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0055http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0055http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0060http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0060http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0060http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0060http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0060http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0065http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0065http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0065http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0070http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0070http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0070http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0070http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0070http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0070http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0075http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0075http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0075http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0080http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0080http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0080http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0080http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0080http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0085http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0085http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0085http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0085http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0090http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0090http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0090http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0090http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0095http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0095http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0095http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0095http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0095http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0095http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0100http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0100http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0100http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0105http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0105http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0105http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0110http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0125http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0125http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0125http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0130http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0130http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0130http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0130http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0135http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0135http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0135http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0140http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0140http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0140http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0140http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0140http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0145http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0145http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0145http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0150http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0150http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0150http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0155http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0155http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0160http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0160http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0160http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0160http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0165http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0165http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0165http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0165http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0170http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0170http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0170http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0170http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0170http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0165http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0165http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0165http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0165http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0165http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0165http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0165http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0165http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0165http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0165http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0165http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0165http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0165http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0165http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0165http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0165http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0165http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0165http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0165http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0165http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0165http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0165http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0165http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0165http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0165http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0160http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0160http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0160http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0160http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0160http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0160http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0160http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0160http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0160http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0160http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0160http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0160http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0160http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0160http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0160http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0160http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0160http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0160http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0160http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0160http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0160http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0160http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0160http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0160http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0160http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0160http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0160http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0160http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0160http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0155http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0155http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0155http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0155http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0155http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0155http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0155http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0155http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0155http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0155http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0155http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0155http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0155http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0155http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0155http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0155http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0155http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0150http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0150http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0150http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0150http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0150http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0150http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0150http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0150http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0150http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0150http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0150http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0150http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0150http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0150http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0150http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0150http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0150http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0150http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0150http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0150http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0150http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0150http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0150http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0150http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0150http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0150http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0150http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0150http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0150http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0145http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0145http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0145http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0145http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0145http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0145http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0145http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0145http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0145http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0145http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0145http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0145http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0145http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0145http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0145http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0145http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0145http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0145http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0145http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0145http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0145http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0145http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0145http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0145http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0145http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0145http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0145http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0145http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0145http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0145http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0145http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0140http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0140http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0140http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0140http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0140http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0140http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0140http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0140http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0140http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0140http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0140http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0140http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0140http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0140http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0140http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0140http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0140http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0140http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0140http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0140http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0140http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0140http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0140http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0140http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0140http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0140http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0140http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0140http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0140http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0140http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0135http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0135http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0135http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0135http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0135http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0135http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0135http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0135http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0135http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0135http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0135http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0135http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0135http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0135http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0135http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0135http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0130http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0130http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0130http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0130http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0130http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0130http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0130http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0130http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0130http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0130http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0130http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0130http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0130http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0130http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0130http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0130http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0130http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0130http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0130http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0130http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0130http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0130http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0130http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0130http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0125http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0125http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0125http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0125http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0125http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0125http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0125http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0125http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0125http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0125http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0125http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0125http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0125http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0125http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0125http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0125http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0125http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0125http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0125http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0120http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0115http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0110http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0110http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0110http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0105http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0105http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0105http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0105http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0105http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0105http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0105http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0105http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0105http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0105http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0105http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0105http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0105http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0105http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0105http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0105http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0105http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0105http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0105http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0105http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0105http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0105http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0105http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0105http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0105http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0105http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0105http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0105http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0105http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0105http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0105http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0100http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0100http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0100http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0100http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0100http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0100http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0100http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0100http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0100http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0100http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0100http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0100http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0100http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0100http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0100http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0100http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0100http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0100http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0100http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0100http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0100http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0100http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0095http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0095http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0095http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0095http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0095http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0095http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0095http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0095http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0095http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0095http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0095http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0095http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0095http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0095http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0095http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0095http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0095http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0095http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0095http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0095http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0095http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0095http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0095http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0095http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0095http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0095http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0095http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0095http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0095http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0095http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0095http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0095http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0090http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0090http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0090http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0090http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0090http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0090http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0090http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0090http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0090http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0090http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0090http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0090http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref0090http://refhub.elsevier.com/S0957-5820(15)00179-2/sbref