EPA-AA-SDSB-86-3


                        Technical  Report
                 Evaluation of the Feasibility
                    of Liquid  Fillneck Seals
                         December 1986
                             NOTICE

Technical  reports  do  not  necessarily   represent   final   EPA
decisions or positions.  They  are  intended to present technical
analysis of  issues using  data which  are currently  available.
The purpose in the  release  of  such reports is to facilitate the
exchange of technical  information  and to  inform the  public  of
technical developments which may form  the  basis for  a final EPA
decision, position or  regulatory action.

            Standards  Development  and Support Branch
              Emission Control  Technology Division
                    Office  of Mobile Sources
                  Office of Air and Radiation
              U.S. Environmental Protection Agency

-------
                       Table of Contents


      Section                                             Page
I.    Introduction/Overview                                2

II.   Evaluation of Liquid Seal Concepts                   5

      A.   Preliminary Analysis and Experimental Design    5

      B.   In-Tube Liquid Trap                             13

      C.   J-Tube Liquid Seal                              17

      D.   Submerged Fill                                 20

      E.   Techniques for Reducing Required Fill Height   23

      F.   Discussion/Conclusions                          28

III.  Bench Testing of a Liquid Seal Onboard System       35

      A.   Description of the  Prototype System            35

      B.   Evaluation of System Efficiency                41

IV.   Conclusions                                         49

References                                                53

Appendix                                                  54

-------
                              -2-
I.   Introduction/Overview

     The  only  significant  uncontrolled  source  of  hydrocarbon
emissions  from  motor  vehicles  is  the  gasoline  vapor  which
escapes   during    vehicle   refueling.    In  . an   uncontrolled
refueling, the vapors  that are displaced during the  refueling
event pass  through the fillneck and  escape to  the  atmosphere.
Systems   designed   to   control  these   emissions   are   called
refueling vapor control systems.  When  the  controls are vehicle
based,  they are generally  referred to as  onboard vapor  recovery
systems.   The   essential   ingredients   of  an   onboard  vapor
recovery  system  are:   1)  a   fillneck  seal to  prevent  vapors
from escaping  to  the  atmosphere  via  the  fillneck,  2)  a  fuel
tank vent and  vapor   line  from  the  fuel  tank to  allow  for
displacement of  refueling  vapors,  and  3)  a method  of  removing
hydrocarbons from  the  vapors  displaced  from  the fuel  tank  and
storing  them  for  purging  at   a  later  time.   Normally  this  is
accomplished through the  use of a  carbon canister.

     This  report  is   primarily  concerned  with  fillneck  seal
portion  of  the  onboard  vapor  recovery  system.   Historically,
the  fillneck  seals  most   often  discussed  as  part  of  onboard
systems  are mechanical seals.   In a  mechanical  seal  system,  a
temporary physical barrier  is  created between  the seal  and  the
nozzle   when the   fuel  nozzle  is  properly inserted   into  the
fillneck  (see Figure 1).   This  barrier  (usually referred  to  as
a seal)  prevents the vapors from  escaping out  the fillneck when
the fuel is dispensed.

     In an extensive development  and  testing  program  conducted
in 1978,  API demonstrated  that  mechanical seals  can effectively
prevent  vapors  from  escaping   through   the  fillneck  over  the
useful   life  of  a   light-duty  vehicle  when used under  normal
conditions.[1]   However,  some  concern has been  expressed about
the  safety,  durability,  and  integrity   of  the  mechanical  seal
approach.

     Concerns  about the mechanical  seal  fall in three areas:

     •      Fillneck seals  may  require  a  pressure  relief  valve
           to  deal with potential fuel  tank overpressures which
           could  occur as  a   result  of failure  of  the  fuel
           nozzle automatic shut-off  mechanism or  a blockage in
           the  vapor line between the fuel  tank and the carbon
           canister.    Without   a  pressure  relief  valve,  tank
           overpressures  could  result in a  fuel spitback  with
           potential  fuel  tank  overpressures.

-------
                      Figure 1
  FILL RPE MODIFICATIONS
  ROTARY SEAL
                   ROTARY SEAL
  TRAP DOOR
           LEAD RESTRICTOR
FILL  PIPE  MODIFICATIONS
ROTARY SEAL
                 ROTARY SEAL
TRAP DOOR
                                          SPOUT
         LEAD RESTRICTOR

-------
                              -4-
     •     Under  some  adverse  in-use  conditions,  a  mechanical
           seal  may  require   maintenance   or   replacement  to
           retain  a  high  in-use efficiency.   This  depends  in
           large part on the material used to make  the  seal and
           the  degree   to  which  the  mechanical  seal  mounting
           within the fillneck protects it from  unusual  wear  or
           abuse.

     •     If not properly designed,  the  mechanical seal may be
           susceptible  to tampering  by the consumer.

     Instead of using  a mechanical  seal  in  the fillneck,  one
alternative  is  the  use of  a  liquid seal.  The  idea  of  using a
liquid  seal  in  the  fillneck  was  introduced  early  in  the
development  of  onboard  vapor   recovery  systems.   A  report
published   in   1973   by   the   American   Petroleum   Institute
identified  and  briefly  evaluated several different liquid seal
concepts.[2]  A liquid  seal  system uses a fillneck  design which
routes  the  incoming  fuel  in  such  a  way  that the  fuel itself
prevents the vapors from escaping via the  fillneck.

     The use  of a liquid seal design would  help  to eliminate
the  potential  problems  associated  with  a  mechanical  seal.
Since a new seal is formed at  each  refueling,  in  use durability
and  maintenance are  not  in  question.   Also, since  the liquid
seal  fillneck  is  no   different  in  appearance  than  current
fillnecks and  the fuel  itself forms the  seal,  liquid seals are
essentially  tamper-proof.   The  problem  of   overpressure during
refueling would be  eliminated,  because  fuel  would  back  up  in
the fillneck as soon  as any  pressure built up.   The fuel nozzle
would then  either  shut off  automatically or could  be  shut off
manually,  since the  nozzle  operator  could  see  the  overfill
condition.    In  addition,  liquid  seal systems  may  be  somewhat
less  costly  than mechanical   seals  since  a  pressure  relief
device may not be required.

     The purpose  of  this  report is to evaluate  the liquid seal
approach  to   fillneck  vapor   containment   as    a   practical
alternative  to  the  more  widely   recognized  mechanical  seal.
After evaluating several different  liquid seal  approaches,  the
performance of  one of  the liquid seal designs is then evaluated
in a simple prototype onboard control system.

     The first  portion  of  the  report  covers evaluation  of the
liquid  seal  concepts.   It  starts with a  conceptual analysis  of
the  liquid  seal and  provides  an overview  of  the  experimental
and  analytical  methodology  used  in  the   testing.   Then  it
examines each of  three  basic  liquid seal designs individually:
the  "in-tube"   liquid   trap,   the  "J-tube,"   and the submerged
fill.   The  testing  done  on  each  design  is  outlined   and  the
strengths and  weaknesses of  each  are  noted.  Next,  a  problem

-------
                               -5-
common  to  each  design,  air entrainment and bubble proliferation
in   the   fillneck,   is   discussed   and   various  methods   of
controlling these  problems  are  evaluated.   The first portion of
the  report closes  with  conclusions  about  the  adaptability of
these   liquid   seal   designs   to   current   fuel  tank  fillneck
configurations.

     The second portion of  the  report  discusses the performance
of a crude,  first  generation  liquid seal fillneck (J-tube) on a
simple  prototype onboard  system.   It  begins with  a description
of the  key components needed for  an  onboard  control  system and
then  describes  the  components used   in  EPA's  system and  how
these met  the performance criteria needed  for  the liquid seal.
The  test   procedure  used  to  evaluate  the  performance of  this
system  in  a  series  of  bench  tests  is described and this  is
followed  by a discussion of  the  test  results  including  system
efficiency and sources of emissions.

     The  report then  closes  with some  conclusions  about  the
performance  of  liquid  seals  and  compares  them  to  results
achieved for mechanical seals.

II.   Evaluation of Liquid Seal Concepts

     A.     Preliminary Analysis and Experimental Design

     1.     Introduction

     The  term  "liquid seal" in the context  of  an onboard vapor
recovery system refers to  any system  of fuel tank and fillneck
in  which  a  column  of   liquid  separates   in-tank  vapor  from
atmospheric   air.    One   key  parameter   in   evaluating   the
feasibility and performance of liquid  seals  is the fill  height
required  to  maintain a  liquid  seal  during  a  refueling  event.
If the  fill  height available  is less  than that needed, the fuel
rising  in  the  neck  will   cause  a premature  shut-off  of  the
nozzle.    One  of the major goals  in   these  evaluations  was  to
determine  the  minimum  fill  heights  necessary  to  completely
refill  the  tank   using  liquid  seals  in  the   fillneck.   This
section develops  some models which help  to  predict  the  manner
in which  these  liquid seals would perform  and  what fill  height
might  be  necessary  to  make  use  of  a  liquid  seal  as  an
alternative to  a  mechanical  seal.  First,  the  static situation
is discussed and a mathematical model  relating  fillneck  height
to  tank  backpressure in   the  static  situation  is  presented.
Then the dynamic effects  that  are  part of  a  typical refueling
event   and   that    complicate   the   mathematical   model   are
introduced.   Third,   an  experimental  design   for  testing  the
various   fillneck  configurations   is   outlined.   Fourth,  the
general  test procedure for  testing the liquid  seal  concepts is
developed.

-------
                               -6-
     2.     Static Situation

     Any  liquid  trap  system can  be  modeled by  the  simple  'U1
tube  system  shown  in  Figure 2.   In  the static  situation,  the
vertical  fill  height,  hf,  required  to  support  any  induced
tank  backpressure,  PB,  can  be found  using the  basic equation
of fluid statics.  In this case:

           PB = Po + pghr, where

     PB  = tank backpressure,  gauge
     P0  = reference gauge pressure at  reference height, h = 0
     p  = density of fluid in tank
     g  = acceleration of gravity at  test altitude
     hf  = required fill  height above  reference height

     For  example,  to support  a  backpressure  of  5  inches  H20
under static  conditions,  a gasoline column (specific gravity =
0.72) would  have  to  rise  approximately  7   inches   above  the
reference height.   So vertical  fill  height would  have to  be at
least 7  inches.   During  testing,  a water manometer was  used to
gauge  pressure  so that  pressure  could  be  read directly  in
inches  of H20.   If  fuels  were   dispensed very   slowly,  this
model could be  used  to  analyze  the system.   It is not, however,
and  the  next  section   looks  at  the  dynamic  effects  of  a
refueling.

     3.     Dynamic Effects

     A  number  of  factors  complicate  this  model  for  a typical
refueling event.  These  include:

     •     Dynamics:   At  a  typical  gasoline   pump  fuel  is
           dispensed   at    a    rate    of    7-10   gallons/minute
           (gal/min);  thus the system is not static.

     •     Injection  effect:    The    kinetic   energy   of   the
           gasoline as  it flows  out   of  the  dispensing nozzle
           tends to aid the  flow  of  gasoline  into  the tank and
           to reduce the required.fill height.

     •     Turbulence  and  air  entrainment:   The  interaction
           between  liquid and  air in  the fillneck   is  quite
           turbulent.    This    interaction   tends   to  greatly
           increase the  required fill  height.

-------
                          Figure 2 - 'U1 Tube
Pb = tank backpressure, guage
Po = o = reference guage pressure
hf = required fillneck height

-------
                               -8-
     Observations  during  informal testing  of  the  'U'  tube  set
up of Figure  2,  support  the intuitive hypothesis  that  required
fill  height  is  directly  related  to  tank backpressure  for  any
given fuel flow rate and fillneck  configuration.   In  the static
situation,  PB  and  hr  are  related  by  a  constant  conversion
factor,   the   liquid  density.   In the  dynamic  situation,  the
relationship   is   much  more   complex,   with   fluid  density,
viscosity, surface  tension,  and  other  factors  all  interacting
in  a  highly   turbulent   flow.    In  this  dynamic,   turbulent
situation,  it   was   expected  that   PB   and   hf  would  vary
directly  and  that  an  increase  in  tank  backpressure  would
increase the fill height necessary to  support  that pressure.   A
mathematical  analysis  of  such  a turbulent,  dynamic  flow  was
outside  the   scope  of  this   investigation,  so  a  mathematical
model  was  not developed.   Instead,  the relationship  between
tank   backpressure   and   fillneck   height   was   determined
experimentally for the three liquid seals evaluated.

     4.     Experimental Design

     The  ultimate  goal of  the tests  conducted  by EPA  was  to
compare  the  performance of  a  number  of  liquid  seal  fillneck
configurations  under uniform  test  conditions  and  procedures.
Because  the   optimal  experimental design was   not  intuitively
obvious  when   testing  began,  a  series  of  developmental tests
were performed with  water  as the test fluid.  These  tests were
done to  evaluate a  number  of  experimental designs without  the
safety concerns and expense involved with gasoline testing.

     These preliminary tests  were performed using  standard  tap
water,  and  fillneck  systems  were  constructed  using  metal,
tygon, and  plexiglass.   These  systems were then  attached to a
standard  automotive  fuel   tank  having  a  nominal capacity  of
about  18  gallons.   All  gasoline  tests  were  performed  with
indolene  clear  as  the  test  fluid,  and  since  plexiglass  is
incompatible  with  gasoline,  systems were  built  of  metal  and
tygon.  The temperature  during this  testing  was the  prevailing
ambient temperature (nominally 80°F).

     The developmental water tests  were  carried  out in  a   'U'
tube system such as the one described above  and diagrammed in
Figure  2.   Data  was  taken   relating   fill   height  to  tank
backpressure for different  fillneck  configurations during these
developmental   water  tests.    The numerical   results  of  these
tests proved  to be  of  little  use in  predicting the  results of
the gasoline testing, however,  because of the great differences
in  the  properties  of  water  and   gasoline   (see  Table  1).
However,   the   tests .  were  very  helpful   in   isolating  the
parameters  that would  prove  critical  in  determining  required

-------
                              -9-
                            Table  1

            Properties of Water and Gasoline at 20°C

Property                        Water    Gasoline (n-Qctane)

Density (Ib/gal)                8.30            5.83


Viscosity (Centipoise)          1.0             0.54


Surface Tension (Dynes/cm)      72.8            21.8
Source:    CRC  -  Chemistry  and  Physics  Handbook  and  Perry's
           Chemical Engineering Handbook, 5th Ed.

-------
                              -10-
 fillneck  height.   The  qualitative  results of  this  preliminary
 testing  which  are  important  to the  experimental  design  are
 outlined below:

     •     The  system performed  as  predicted  by  the  'U1  tube
           model  -  required  fill  height  varied directly  with
           tank backpressure.

     •     Increasing  fuel  inflow rate  increased  required  fill
           height.   Bubbling  caused   by  the  mixing  of  fluid
           entering  the fillneck with  air in  the  fillneck was
           observed to increase as dispensing rate increased.

     •     Changes  in  fillneck  diameter  (or   volume  per  unit
           length)  affected  required  fill  heights.   Generally,
           required   fill  heights  increased   with   decreasing
           fillneck diameters.

     •     The  orientation  of  the  dispensing nozzle  in  the
           fillneck  affected fill height.   If  the  nozzle  was
           oriented  appropriately, the  liquid  tended  to adhere
           to  the  fillneck  wall.     If  this   was   done,   air
           entrainment  and  bubbling  were  reduced leading  to  a
           decrease in required fill  height.

     The  results  of  these  tests confirmed  that  these  factors
would have to be controlled to ensure consistent test results.

     Based on observations during the  water testing, the U-tube
system  appeared to  create  a  very  adequate  seal.   To  further
evaluate the  feasibility  of  this approach, a "u" was built into
a  2-inch  plexiglass pipe and  tested  using water.   As  will  be
discussed  below,   the  testing   showed  that  the  smaller  cross
sectional  area  for  flow  within  the  trap  required  too high  a
fill height to  be  practical.   However, this experiment revealed
useful information on flow and turbulence characteristics.

     Based  on  the  experience  gained  from  this  preliminary
testing  and  supplemented by information  contained  in  the SAE
Recommended  Practice  regarding   refueling  emissions,  standard
test conditions were chosen to  test  the  liquid seal  fuel  tank
systems using gasoline  under  conditions  likely  to  occur during
a  typical  refueling  event.[3]   These conditions  were  used in
the  tests  which  evaluated  the   fill  height   requirements  and
other performance aspects of the liquid seals system.

     The  backpressure  was  set  at   five   inches  H2O.   It  is
expected that  carbon  canisters  and   their accompanying  tubing
and  valves  can be   designed  to accommodate   normal  refueling
rates with  tank backpressures of 5  inches H2O or  less.   This
is  supported  by  experimental  evidence  presented  in  an  API

-------
                              -11-
report dated October,  1978.[1]   The  report quotes backpressures
as  low  as  2  inches  HZO,  at  a   refueling   rate   of   eight
gal/min.   All  three of  the fillneck seal  approaches  evaluated
by EPA were  tested  at  5 inches  H20  backpressure and  most  were
tested at some lesser backpressures.

     Fuel  dispensing  rate  was also  standardized.   Typical  fuel
dispensing  rates  average  from  7 to  10 gal/min,  but  the  fuel
cart  available  for use  in  this  testing  had  a  maximum  fuel
dispensing  rate  of  7.5  gal/min.    Therefore,   this  was   the
designated fuel flow rate for gasoline testing.

     As was mentioned  above,  fillneck cross sectional area  also
affects  required  fillneck  height  and  had to be  standardized.
Because  most   stock  fillnecks  have   an  outer   diameter  of
approximately two inches,  this was  chosen  as  the  base fillneck
diameter and tygon or metal tubing of this  diameter was used.

     Also  mentioned  above,   fuel  nozzle  orientation   can  also
affect  required  fill  height,  but  the  optimal  orientation  is
different   for   each  fillneck  configuration.    Also,  nozzle
orientation  is  one  of  the  key   human   factors   in  vehicle
refueling,  and may  vary  significantly from operator to operator
and  vehicle  to  vehicle.   For   these  reasons,   fuel  nozzle
orientation  was  not  standardized  to  minimize  fill  height.
Instead,   for  each   fillneck  configuration  the   nozzle   was
oriented in  the  manner that would normally  occur  in-use  if the
nozzle were to rest vertically in the fillneck perpendicular  to
the surface.

     The  standard  test  conditions  described  above   were  then
incorporated into the standard test  procedure  developed for the
gasoline testing.   The  standard setup  for  testing the  liquid
seal designs is shown  in Figure 3.   The gate valve  mounted  on
top of  the tank was  used  to regulate  the size of  the orifice
through  which  any   vapor   leaving  the  tank  must  pass.   The
backpressure  induced  in the  fuel tank was controlled for any
fuel inflow rate by changing the position of the gate valve.

     Prior to the start of each  test, the  manometer was zeroed,
the fuel  tank was  leveled,  and  the  gate  valve  opened to about
one fourth of wide  open.   The fuel cart was  turned on and  fuel
dispensed  at  the highest  available  rate.   The gate  valve was
then closed  until  backpressure  was  stabilized  at  the desired
value   (initially,   5   inches   H20) .     The   height  of   the
fluid/bubble  column in  the  fillneck  was  then  measured  when
backpressure  stabilized.   This  value was  recorded  as  the
required fill  height.   The tank  was drained  and  the  operation
repeated.

-------
                                         Figure 3


                                Standard Experimental Setup

                               h = required fillneck height
                                      Fill Nozzle
Box Represents
Liquid Seal
  System
                                                              Gate Valve

-------
                               -13-
      At  this point,  fuel RVP and dispensed  and tank temperature
 were not important.   Neither  the RVP of  the fuel or  the  fuel
 temperatures mentioned above were controlled.

      The testing of each  system generally  adhered  to  the  test
 procedure  outlined above,  although  some  details  were  specific
 to  the  individual  systems being tested.   The most significant
 difference  between  individual  tests  was  the measurement of  fill
 height.  Fill height can be generally defined  as follows:

      The value  of  the vertical  component  of  a   vector  from  a
      point  on the plane  of full tank  fuel level to  the point
      where   the   tip   of   the  dispensing  nozzle   rests  in   the
      fillneck (the vertical distance  from the tank  full level
      to  the nozzle tip).

      The peculiarities of  the measurement  of fill  height   for
 each individual  system will  be  discussed  in  the  sections of
 this report  which  deal with  the specific  designs.   It  should
 also be noted here  that  fill height,  as used in this  report,
 does not include  the  distance from the  inserted  nozzle  tip to
 the  gas  cap location.   This distance (2-3 inches)  would have to
 be included in any fillneck design.

      B.     In-Tube Liquid  Trap

      1.     Introduction

      The first  liquid seal approach  tested  was  the  "in-tube"
 liquid  trap.   The  in-tube  trap was  chosen for study because  the
 preliminary water testing  showed  that  the system could  provide
 a  positive   seal  and  could be installed  by modifying  only  the
 fillneck,  leaving the  tank essentially unchanged.   If  this  were
 possible,  and the fillneck height  requirements  permitted,   the
 design   could  be  adapted   to  fit a number of  vehicles.   This
 section  looks at the in-tube trap  and  its performance under  the
 standard test conditions.

      An  in-tube  trap   in  the  fillneck  operates   by  creating  a
 liquid  barrier between the gasoline vapor in the tank and  the
 atmosphere.   Its operation is similar  to that of the home  sink
 drain pipe,  but  the 'U1 shaped trap is  built into the  fillpipe
 rather  than bending the fillpipe  into a  'U'  shape.

      A  diagram  of the in-tube  liquid  trap and illustrations of
•the  parameters describing  its  construction  are  shown in  Figure
 4.    The trap  was  designed  to  divide  the  tube  into three
 channels of equal cross-sectional  area.   Liquid  is  trapped in
 the   upward  opening   'U',   and  if   tank  backpressure  increases

-------
                  Side View
          Figure 4

              In-Tube Trap
Bottcm
  View
                    e-b
             T
               At
— 11
                                                                   Dimensions
                                                                   Plexiglass

                                                                   a = c = 0.73"
                                                                   b = 0.54"
                                                                   D = 2.0"
                                                                   Metal

                                                                   a = c = 1.1"
                                                                   b = 0.8"
                                                                   D = 3.0"
                                                                    d  =  Trap Depth
                                                                    A = B = C

-------
                              -15-
during  refueling,  a  liquid  column  will  rise  above  that  'U'
until  the  pressure  is  offset.   "Trap  security" describes  the
susceptibility  of  the  trap  system  to  allowing  vapor  to  pass
through  it  during the  refueling  event.   Trap  security  can  be
increased  by  increasing  the  depth   of  the   liquid  seal,  but
increasing this depth might  also increase  the required fillneck
height.  Normally, with an in-tube liquid  trap,  the  depth would
just  need  to  be  adequate  to  accommodate  refueling on  sloped
terrain.   So,  to  refuel  on  a  30°   incline,  a  trap  depth  of
one-half inch  would  be sufficient (for the  large trap  shown  in
Figure 4).

     Two traps  were  built  for  testing.    The  smaller trap  was
set  in  a  two  inch  diameter  plexiglass  tube.   This trap  was
built so that flow through the  trap  could  be visually monitored
during the preliminary  water  testing.   A larger 3 inch diameter
trap was built entirely of metal so  that gasoline could  be used
as the primary test fluid.

     2.    Test Procedures

     Figure  5  is  a schematic of  the  experimental setup  used  to
test  the in-tube  liquid  traps.   It is important  to note  the
location of  the  trap  in  relation to  the  tank  full  level.   If
spatial    efficiency    were    optimized    in    a    production
configuration,  the trap would  actually  be at  the  roof  of fuel
tank inside  the  vapor  space  and,  the upward  opening  'U'  would
open  at  the  tank  full  level.   However,  for  the  sake  of
convenience,   these  traps were  built  on  top  of the  tank,  and
hence  required fill  height  was  measured  from  the  top  of  the
upward opening  'U' to the nozzle tip.

     3.    Results/Discussion

     As  was  mentioned  previously,  some   of   the  developmental
testing  discussed earlier in  this report was done on an in-tube
liquid  trap.    The  dimensions  of that  trap  are those  listed
under  "Plexiglass"  in  Figure  4.    The  developmental  testing
showed that  a trap built  inside a  2  inch diameter  tube would
not  be  large  enough  to  support a  backpressure  of  5  inches
H2O.  The  channels  through which liquid  was  supposed  to  flow
(approximately  1  square inch)  simply  were not  large  enough  to
allow liquid to pass  through quickly.

     In order to accommodate gasoline  and  higher flow rates and
to  reduce  required fillneck  height,  a larger  in-tube  trap was
set in a 3  inch diameter  tube.  The  dimensions  of  the trap are
shown  in  Figure   4  under  "Metal".    Using  the standard  test

-------
                                    Figure 5

                       Experimental Setup For  In-Tube Trap
                                          Fill Nozzle
Required
Fillneck
Height
Full Tank
  Level

-------
                              -17-
procedure, the tank  was  filled with  a  required fill height  of
only  11  inches.   This is  a  very good  result,  considering  that
in  the  static situation  a neck  height of  7  inches  would  be
needed  to support a  5 inch  H20  backpressure  (5  inches/0.72  =
6.94  inches).   The  implications  of  this  result  are  discussed
later in the report.

     4.     Conclusions

     A  fill  height of approximately  11 inches  for  the  in-tube
liquid trap  is quite  acceptable  for many  vehicle  applications.
However,  the in-tube  liquid  trap  approach has  some drawbacks.
First,  the  in-tube   trap  could   be  relatively   complex   to
construct on a production  basis.   Also, since the trap required
a  3  inch  diameter  fillneck  to  perform  adequately under  the
standard  test  conditions,  current  fillpipe designs might  have
to  be modified  to accommodate  its use.*   A  3  inch  diameter
expansion  in  some   fillnecks  may  create  vehicle  packaging
problems.  However no  fuel tank  changes would  be  necessary  and
the  in-tube  liquid trap has  all  the advantages  of  liquid  seal
systems  that were  described above.   Packaging  problems  could be
eliminated if  the in-tube trap were built into  the inside  top
of  the   fuel  tank.   The  advantages  and  disadvantages  of  the
in-tube  trap are  discussed further  following  the examination of
the other liquid seal approaches.

     C.     "J" Tube Liquid Seal

     1.     Introduction

     The  second   liquid   seal   approach   to   refueling  vapor
containment  examined was  the  'J'  tube liquid seal.    The  'J'
tube  was  chosen  for  study because  it  was  a  logical  extension
from  the  in-tube  trap  approach  and  would  take  only  minor
changes   in  fuel  tank  and  fillneck  to  install  and has  all  the
advantages  of  liquid  seals.   This section  looks at  the  fil'l
height  necessary  for the  'J'  tube  system  when evaluated  under
the standard test conditions.

     The  'J' tube  system  is  shown in Figure 6.  The 'J' tube is
simply a  modified  'U'  tube  (like that  shown  in Figure  2).   The
difference  is  that   in  the   'J'  system  the  fuel  tank itself
replaces  the downstream,  vertical  portion  of the  drainpipe.
     The ultimate size  of  the trap and fillpipe diameter needed
     depend  in  part   on   the  system   backpressure.    System
     backpressure  less  than  the  5-inch  H2O  used  here  are
     achievable, and  would reduce the  diameter of  the  in-tube
     trap.

-------
Figure 6 - 'J1 Tube

-------
                              -19-
Th e dimensions  of  the  'J1  system  along with  illustrations  of
some of  the  descriptive  terms used in  the  discussion are given
in  Figure  6.   it  should  be noted  that  the  liquid trap  is
located  inside  the  fuel  tank,   and  the  top  of  the trap  is
located on or  near  the full  tank plane.   This  promotes spatial
efficiency and makes maximum use  of  available fillneck height.

    - 2.    Test Procedure

     Because   the  'J'-tube  trap  is  located  inside of  the  fuel
tank and  fill height  is  measured  from  full  tank  level  to  the
nozzle  insertion  point,  the  standard  test method  was followed
exactly in this case.

     3.    Results/Discussion

     The 'J'  tube  test results are listed below:

              Fuel dispensing rate = 7.5 gal/min.

            Backpressure              Required
            (inches H20)
                5.0                     16.0
                4.0                     13.5
                3.0                     11.0

     The 16  inches  of  fill  height required to  fill  the tank at
five  inches  H20 backpressure  was more  than  expected  based on
extrapolation of  the results  of  the developmental  water  tests
discussed  earlier.   Although  bubbling,  due  to  the mixing of
water   with   air   in   the   fillneck,   was    observed  during
developmental  tests,  the   violence   of   the  bubbling  during
gasoline   testing   was   not  fully   expected.    The   extensive
bubbling in  the fillneck was  the cause  of  the  large  required
fill heights.   It  was  clear from this  testing  that  if  bubbling
could   be  controlled,   required  fillneck   heights  could  be
reduced.  This cou-ld be  managed  by  reducing the backpressure or
controlling the proliferation of  bubbles  in the fillneck during
refueling.

     4.     Conclusion

     The 'J1   tube,  a simple modification of the  'U1  tube, was
tested  using  the  standard   test  method.    Excessive turbulence
and  bubbling  in the  fillneck  made a  fill  height of  16 inches
necessary  for  refueling  under  the  standard  test  conditions.
The  control of  the  bubble problem,  and  some methods  of  fillneck
height reduction are discussed  in section E.

     It is important to  note that  the  'J1  tube  trap  appeared to
be doing its  job of controlling  emissions.   A rough  measurement

-------
                              -20-
of  the  hydrocarbon  concentration  at  the  fillneck mouth  was
taken  with  a  lower  explosive  limit  (LEL)  meter.   The  meter
showed no  difference  between  the  readings taken at the fillneck
mouth and  readings taken around the test chamber.

     D.    Submerged Fill

     1.    Introduction

     The  third  and  final  liquid seal  system  investigated  was
the submerged  fill.   In a  submerged  fill  system,  the  fillneck
opens into the  fuel  tank below  the liquid  surface.   By  filling
the tank  from  below  the surface,  a   liquid  seal  is  naturally
formed between in-tank vapor and atmospheric air.

     A   drawing  of   a   submerged  fill  system   is   shown   in
Figure 7.   The  significant  features  of  the  submerged  fill
system are listed below:

     •     The  security of  the  liquid  trap  is  good  and  it
           improves as the tank fills.

     •     A   pressure   relief  valve   would   have   to   be
           incorporated into any  submerged  fill system.   If  the
           lines  from  fuel  tank  to  carbon  canister   became
           blocked, tank  pressurization during refueling  could
           become  a  problem.    If the backpressure  in the  tank
           rose  between  refuelings,   when   the   gas  cap   was
           removed  from the  fillneck,  the  overpressure  could
           cause  gasoline  to  spit  out   of  the  neck.    The
           pressure  relief  valve  is   needed  to  eliminate  this
           problem.

     •     The fillneck height  is not  critical  during  the early
           stages of  refueling.   Only in the last 10 percent of
           the fill is  the  full  fill  height  used,  as the  back
           pressure  caused   by  the  fuel  level   in  the  tank
           increases.

     2.     Test Procedure

     In a submerged fill system, the location of  the top  of  the
liquid trap  rises  as  the tank fills.   One major change was made
in  the   standard   test  procedure  to  accommodate  this   unique
feature  of the submerged  fill.  Since  the  top  of  the  liquid
trap rises during refueling, the  entire  fillneck  height  is  only
needed during the  final  stage  of  refueling; when  the top of  the
trap approaches the full tank liquid  level.   Therefore required

-------
Figure 7 - Submerged Fill

-------
                              -22-
fill  height  must  be  measured  just  prior  to  the  end  of  the
refueling  event.   For  these  tests,  fill  height was  measured
when  about  18  gallons  of  fuel  had  been  dispensed   (into  a
nominal 18 gallon tank).

     3 .    Results/Discussion

     The  results  of  the  testing of  the submerged  fill system
are presented below:

                    Flow rate:   7.5 gal/min

            Backpressure              Required Fill
            (Inches H20)              Height (inches)

                5.0                         16
                4.5                         13

     These results are similar  to  the  results for the  'J' tube,
and the bubble problem was evident once again.

     There has  been  some previous  work  done  to evaluate  the
effectiveness  of   the  submerged  fill   seal   in  controlling
refueling  vapors.   Preliminary  lab work  done by EPA  in  1979
found the  submerged  fill  to  be  an  effective method of refueling
vapor  control.[4]   Crude  sampling  done  with  an  LEL  meter
supports  this  finding.   Samples  taken  during  refueling  from
near  the  mouth  of  the  fillneck  showed no  higher  hydrocarbon
concentrations than did background samples.

     4.    Conclusions

     The  submerged   fill  approach   was  evaluated   using  the
standard  test procedure,  and required  fill  height was  measured
just  prior  to  completion of  the  refueling.  At  a backpressure
of  5  inches  H20  a fill  height  of about  16 inches  was needed
to  accommodate  the  large  liquid/bubble column in the   fillneck
caused  by  excessive  turbulence  and  air  entrainment  during
refueling.

     There are  fuel  tank  fillneck configurations currently  in
use  that  could not  be easily  adapted  to  accommodate   fillneck
heights of  16  inches.   Since  bubbling  in  the fillneck is  the
reason for the larger  fill heights,  a  reduction in the bubbling
should lead  to  a  lower required fill  height.   The next section
of  the  report   addresses the  bubbling problem and  possible
control techniques.

-------
                              -23-


     E.    Techniques For Reducing Required Fill Height

     1.    Introduction

     All  the  liquid  seal  systems built  with  2  inch  diameter
tubing  needed  extended  fillnecks  to  accommodate  the  large
gasoline/bubble column that rose  above  the liquid/air interface
in  the  fillneck during  refueling.   This  section of  the report
examines  bubbling  and  air  entrainment  during  refueling  and
evaluates  some  possible  methods  of  fill  height  reduction.
Bubbling  is caused  by:   1) the  violent mixing of  the gasoline
leaving   the   dispensing  nozzle  with  that   already  in  the
fillneck, and 2) air entrained in the fluid  flow  by the venturi
system in the dispensing nozzle.

     There are  three  basic  means of  controlling bubbling in the
fillneck through modification of the  fillneck.  The first is to
reduce the  violence of  the mixing  of  dispensed  fuel with that
already in the  fillneck.   Another is to  physically obstruct or
direct  the  backup  of  bubbles  in the  neck.   The  third  is to
create  a  system  to  dissipate  the   kinetic  energy  of  the
dispensed fuel.

     Also, part  of  the  bubble  problem could  be  eliminated  by
reducing  the  amount  of air being  entrained in the fuel during
refueling.  This  air  entrainment  is  caused by  two  phenomena.
First, the  automatic  shutoff  venturi  system  in standard fuel
dispensing  equipment  adds  air   to  the  fuel   being  dispensed.
Second, air  is entrained  in  the  fuel  as  a by-product  of  the
turbulence in the fillneck.  The second  of  these two phenomena
can be affected by fillneck modifications  while the first could
only be changed  by  modifying  the nozzle.   This section looks at
the problems of air entrainment  and  bubbling,  and examines some
fillneck modifications  designed  to control bubbling  and reduce
fill heights.

     2 .    Air Entrainment

     The  entrainment  of  air  in the   fluid  flow  affects   fill
height in two ways.   First,  as fuel is dispensed and mixes with
fuel in the fillneck, air is drawn into the  mixture and bubbles
are  formed  increasing  required  fill  height.   Second,  if  air
makes its way into  the fuel tank with  the fuel, the amount of
vapor  forced   out  of  the  tank  during refueling  and  the tank
backpressure  (for  a  given  flow  rate)   are  increased.   The
bubbling  problem  is  discussed  in  the   next section   of  the
report.  This section looks at the entrained air  that makes its
way into the fuel tank.

-------
     A test was  performed  on the  'J'  tube system to  determine
if air was entering  the  tank along with gasoline.   The tank was
filled with 20 gallons of  gasoline and the vapors displaced  by
the  gasoline  were  collected and  the  volume of  those  vapors
measured by pulling  the  vapors  through a dry gas meter.*   Both
the dispensed and tank fuel  were  at the prevailing  temperature,
so vapor  expansion  or  shrinkage   should  have been  near  zero.
The results of the tests  are shown below:

                    Gasoline Dispensed      Vapor Collected
      Trial No.      	(gallons)              (gallons)

          1                  20                    22.2
          2                  20                    22.4
          3                  20                    22.8
     No  attempts were  made  to  identify  the  sources  of  the
entrained air, but it was  fairly evident that it was  caused  by
air bubbles  being  trapped  in the fuel entering  the  tank.   This
in  turn  was  caused  by  the  turbulent mixing  in the  fillneck.
Although this testing was  conducted  on the J-tube system,  it is
also likely  to occur  to similar  degrees  in both the  submerged
fill and in-tube  trap.

     The phenomenon of  entrained air entering the fuel  tank is
significant.  When the  carbon canister  and tubing are designed
for  a  given  liquid seal  configuration,  the  amount  of  air
entrained  must   be  determined   and   figured  in  the  design.
Systems that  allow more air  to  enter the fuel tank will need to
transport more  vapor  at  a  given fuel  inflow  rate.  This  may
also  increase  the  hydrocarbon  load to   be  captured  by  the
canister, since  the  entrained  air may  enhance  evaporation  of
the dispensed fuel.    Some  of  the entrainment  may  be  avoided
through  the  control  of  the  mixing  of gasoline  and  air in  the
fillneck.  This  is examined next.

     3.    Bubbling and  Turbulence

     The most obvious source of  bubbling in the  fillneck is the
violent  mixing,  of  the  gasoline  being   dispensed  with  the
gasoline already in the fillneck.   To examine bubble formation,
a tygon  fillneck was  mounted on the  'J'  tube  system,  and  the
pattern of  bubble  flow  in  the neck was  observed.  Although the
pattern  of   flow  depends   on   the   orientation  of  the   fuel
     It was  possible  to put  20 gallons  of  liquid  in  a  tank
     having a  nominal  capacity of  18  gallons because  the  tank
     was modified with a  liquid  seal  and vent, but did not have
     a  fill  limiter.   Thus  the  vapor  space  of  the  tank  also
     contained fuel,  which would  not normally be the case.

-------
                              -25-
dispensing nozzle  in  the fillneck and  on  the configuration and
cross-sectional  shape  and  area  of   the   fillneck,  a  typical
bubble pattern  is  shown in  Figure  8.  The  pattern shown was
generated   in   a  • two  inch   diameter   tube   with   a    round
cross-section.   The sketch  is  helpful  in  that it highlights the
point of bubble generation  and shows  the  route  that the bubbles
take  in   climbing  toward   the  automatic  shutoff  port  on the
nozzle.     Since   gasoline   bubbles    in   the   fillneck    could
contribute  to  premature  nozzle   shut-offs,  several  different
techniques  to  reduce  bubble  generation  were  examined.    These
included approaches  to eliminating bubbles before  they reached
the  fuel nozzle  and approaches to reducing bubble  formation by
decreasing  the  amount   of  turbulent  mixing  in the  fillneck.
While   the  approaches   examined   here    were   by  no    means
comprehensive,   (i.e.,  many  other options  exist)  one  approach
identified was  quite successful.

     This approach  involved  replacing  a section of  the  standard
2  inch   diameter   tubing with  a  section  of  4 inch  diameter
tubing.   The resulting  'reservoir1 added  volume to  the  fillneck
and  served  as  a  settling  chamber in  which  the  bubbles  formed
and  burst  without  rising  in the  fillneck.   The use of  such  a
reservoir,    the   dimensions   of  which   were    chosen   rather
arbitrarily,  had  a  significant  impact   on  required  fillneck
height.       -

     The reservoir  idea  was  tested as  an  addition to the j-tube
setup discussed  in Section  C.   The  reservoir  is diagrammed in
Figure 9.  The procedure used to gauge required fillneck height
was  the   same   as  that  used   for   the   J-tube  without  the
reservoir.   The results of  testing this system are shown below:

                Fuel  inflow rate  = 7.5  gal/min

            Backpressure           Required Fillneck
            (inches f^O)            Height  (inches)

                5                       11
                4                        9.5
                3                        6.5
     Similar  results  were  obtained  when  the  in-tube  trap and
submerged   fill   systems   were   tested  with   reservoir   type
fillnecks.  The  in-tube  trap set in 3  inch  diameter tubing has
a  "built  in"  reservoir  - the tubing  is larger  than  the  stock
system already.   This  seal system needed a  fill  height of only
11 inches.  The  submerged fill system,  with  the  reservoir used
for the  'J1  tube attached, achieved similar  results (11 inches
required  fill  height).   The  results  of  this  test  show  that a
properly  designed  reservoir system can  greatly  reduce required
fillneck height.

-------
  Figure 8
Bubble Flow
   Pattern
             Static Case
               Liquid Level

-------
                                 Figure 9



                                 Resevoir
                                             Top View
Side View
                      To Nozzle
                         2. in.-
                   «	   4 in.
                                       Stock Fillneck
                                                  Reservior
                          To Tank

-------
                              -28-
     Another  approach  to bubble control  involves  the use  of  a
simple baffle  to  control flow in the fillneck as is now used in
some vehicle designs.  More specifically  the  inflow  of  fuel and
the  backflow  of  bubbles  can be separated  with the  use  of  a
baffling system.  The simple  baffling system  shown  in  Figure 10
directs  and smooths  the inflow of  fuel  and also  directs  and
restrains the backflow of  bubbles.   The  baffle  restricts inflow
to  one  portion  of  the  fillneck  and  encourages  much  of  the
bubble backflow to  rise  in the  other.   The  vents near  the top
of  the  baffle allow circulating  air and  vapor  to  pass  through
the baffle,  but  any remaining bubbles  are popped by  the  edges
of the vents before they can  reach the nozzle venturi tube.

     Tests  were  conducted  to  determine the  fillneck  height
requirement  for  the  submerged  fill   tank  with   a   baffled
fillneck.   The  procedure  described in  Section  D.3.   for  the
submerged fill without baffling was  followed.  At  a  fuel inflow
rate  of  7.5  gal/min  and  tank  backpressure of  5  inches  H20,
required fillneck height was  reduced to  11.5  inches,  a decrease
of 5.5 inches.

     The  baffle  approach  was  not   evaluated with  the  in-tube
trap or  J-tube,  but could  be used  if  the baffle  extended for
enough into the liquid to separate the liquid/bubble flow.

     4.    Discussion

     The  reservoir  and  baffle approaches discussed  above  were
successful in reducing fillneck  height  requirements  by about 30
percent.   Of  the portion  of  the  fillneck  height  that can be
attributed  to  the dynamic  affects  of  refueling,  the  reduction
is   approximately   55   percent.    This   reduction   is  quite
significant given  the  limited time  and resources  committed to
this effort, and suggests further improvement is possible.

     F.    Discussion/Conclusions

     1.    Summary

     One of  the  key components of  an onboard  refueling  vapor
recovery system is  a  fillneck  seal to  prevent  the  escape of
vapors during  refueling.   While  mechanical seal approaches have
been demonstrated successfully,  there  are  concerns  about  the
safety,  durability,  and   integrity  of   the  mechanical  seal
approach.   As  an  alternative  to  a  mechanical  seal,   EPA has
evaluated  the  practicality   of  implementing   a  liquid  seal
approach.

     The  practicality   aspect   of   the   liquid  seal   approach
involves  three  factors:   fill height requirements,  safety, and
system efficiency.   System  efficiency  is  discussed  in  the  next

-------
                              Figure 10



                          Baffled Fillneck
                                 Nozzle
Side View
Front View

-------
                              -30-


portion of this  report.   Fill  height  requirements are important
because of  concerns  regarding  premature nozzle  shut-off.   In
order  to  evaluate the  potential  suitability of  various  liquid
seals  with  respect  to  fill height  requirements, systems  were
tested under  standard test conditions:  a  fuel dispensing rate
of 7.5 gal/min and  a  tank backpressure of  5  inches H20.   The
vertical  fillneck  height  required for  normal  refueling  on each
at the three  liquid  seal  approaches  was measured and  recorded
for each system.

     Due  to  the excessive  amount of  bubbling generated  during
refueling,  fillneck  heights of at  least  16 inches  were  needed
to   refuel  the   tank  for   the   J-tube   and  submerged  fill
approaches.   Because  of fillneck  height limitations  imposed by
some  current  vehicle models,  efforts  were made  to  reduce the
fill  height  requirements  of  the  liquid   seal  systems.   This
involved  fillneck   modifications  aimed   at  controlling  air
entrainment and  bubbling  in the  fillneck.   Through the  use of
reservoirs  and  baffles,   fill   heights  were  reduced  to  11
inches.  The reduction  of  5 inches is  significant,  considering
that  seven  inches  of  fillneck   height  would be necessary  to
support  a   backpressure   of   5   inches    H20   under   static
conditions.   The  effect  of  dynamics  was  reduced to 4  inches.
The  results of the  fill  height  evaluations  are  summarized in
Table 2.

     The   safety   of  liquid  seal   approaches   must   also  be
considered.     Most   notably,   liquid   seal   systems   cannot
pressurize during  refueling such as  could  occur  for mechanical
seal systems.   Any  increase in pressure during refueling would
cause  fuel  to rise  in the fillneck and  automatic  shutoff  to
occur.  The one  safety  problem that  must  be  addressed  for the
liquid seal  system  is  the possible  spitback of  fuel when the
gas cap is removed prior to refueling.

     Spit back could occur  if  the lines to the carbon  canister
became  blocked  and pressure  built  in   the  tank  prior  to
refueling.   When  the gas  cap was  removed,  some  of  the fuel
forming the liquid  seal could be forced out  of the  fillneck as
pressure  equalized.   Spit  back  can  be avoided  by  either  1)
removing  the   liquid seal  after  refueling  or  2)  including  a
pressure  relief  system  to  vent tank  pressure before  spit back
can  occur.    The  in-tube  trap  and   'J'   tube   use the   first
alternative.    By  drilling  a  small hole  at  the  bottom  of the
upward  opening  'U'  in  these designs,  the  liquid  trap can
operate during refueling,  and then drain  soon after  the  event
has been  completed.

-------
                              -31-


                            Table  2

               Bench Test Fill Height Requirement
System              Static

In-Tube Trap          7"

J-Tube                7"

Submerged Fill        7"
 Base
System
 16"

 16"
    Modified
     System

11" (3" pipe diam.)

11" (reservoir)

11" (baffle)
5"  H2O  backpressure,  7.5  gpm  dispensing  rate,  2"  fillpipe
diameter.

-------
                              -32-
     The submerged  fill  system,  on the other hand, must include
a more  complex  pressure  relief  system.   There  are a  number  of
ways  to   incorporate  the   pressure  relief  system  into  the
submerged  fill.  One of  the simpler pressure relief  systems  is
shown in  Figure 11.  The  small  tube  enters  the  tank above the
full tank  liquid level,  so  it  is never submerged.   As refueling
takes place,  the 'U1  portion  of the  small  tube  is filled with
fluid.  When the gas  cap  is  removed  prior  to  refueling,  the
pressure is  vented  through  the  small  tube before  spitback can
occur.  The amount of pressure needed  to  vent the  system can be
changed by changing the depth of  the  'U' portion  of the small
tube.  Although  the spit  back problem can be easily  solved for
any  of  the  systems,  the  alteration of  the  submerged  fill  is
clearly more complex than that required for  the  in-tube trap or
'J' tube liquid seals.

     2.     Conclusions

     There are  two  main  questions  to be addressed in concluding
the  first  portion  of this  report.   First,  is  the  liquid seal
approach to  fillneck  vapor  containment  a practical alternative
to  the  mechanical  seal  as  part  of  an onboard  refueling  vapor
recovery system?  Second, which  systems tested in the study are
the most promising for further development?

     The answer  to  the   first  of  these questions  is generally
yes.  The  liquid seal could be  adapted  to  fit  most light—duty
vehicles and  trucks as  currently  produced,  but  there  are some
vehicles that do not  provide adequate fill height to employ the
liquid  seal.    There   are  two   major  classes   of  fillneck
configurations,  the side  fill  and  the rear fill.  The side fill
configurations will  be discussed first.

     The  liquid seal  design  is  adaptable  to  most   side fill
vehicles.    In  fact, most  vehicles  could probably be equipped
with  an unmodified  fillneck  (no   baffle or  reservoir).   For
other vehicle models,  the  length of  the  fillneck is sufficient
to  permit  the  use  of a  liquid seal  with  a modified,  bubble
reducing fillneck.   Further,  it appears  that changes  could be
made in most side fill vehicles  to  increase  the  fillneck height
if  needed, by .changing  the  location of  the  gas  cap  on the
vehicle body.

     Rear  fill  configurations are  much  less  receptive  to the
liquid seal  approach  than  are side  fills.   The  fillneck height
of  some  rear  fill vehicles  is as   little as 5 inches.   A fill
height  of  5  inches could  accommodate a backpressure  of only
3.15 inches  H20 under static  conditions.  The  limited testing
conducted  by  EPA  suggests  that   with  a  carefully  designed

-------
                        Figure  11

           Submerged Fill with  Pressure Relief
,Tank  Overpressure
        Relief Valve
                                             Full Tank Level

-------
                              -34-
reservoir,  a  fillneck height  could  probably  be  reduced  to
approximately  6.0  inches  (3.0 inch  H20  backpressure and  a  7.5
gal/min  fill  rate).   If  a  manufacturer  were  convinced of  the
superiority of the rear fill  configuration,  and wanted to use a
liquid seal system, changes in  fuel  tank and gas  cap  locations
would  have  to  be  made.   However,  these  changes  might  be
relatively  involved.   Another  approach  would  be  to  create  a
horizontal  reservoir   by  extending  the  fillpipe  horizontally
across the  top of the  fuel  tank,  (either  outside  or  inside).
The  fuel  would  then  enter   the  tank  at  the  opposite  end.
Another  option open  to  manufacturers  of  rear  fill  vehicles
would be  the  use  of  a mechanical seal.   The accompanying safety
concerns  would  have   to  be  addressed.    Nevertheless,   it  is
important  to  note that  most  current  rear fill  vehicle models
are being phased out  and no  new models of  this  design  have been
introduced  recently.    In  the  post  1990 time frame  few,  if any,
new  models  with   rear  fill  tanks  are   expected  in  the  new
car/light truck fleet.

     The  second  question  to  be   addressed  is  which  system
examined  is promising  for further  development.   There  are four
characteristics  that   are  important  in  comparing  these liquid
seals.   The  first  is   the spatial  efficiency of  the system  as
measured  by  required  fill  height.   The  test  results  suggest
that  all  of the systems  can  be  modified  to  get  nearly  equal
results.    Each system  evaluated  needed   a  fillneck height  of
approximately  11  inches  to  support  a  backpressure  of  5 inches
H2O  at  a  fuel dispensing rate  of 7.5 gal/min.   The  submerged
fill system does have  an advantage in this  area,  however,  since
the  full  fill  height  is needed only as the  tank approaches the
full level.

     Refueling safety is  the  second major  characteristic  for
comparison  of  the  liquid  seal  systems.   Both  the  in-tube trap
and  'J'  tube  need  only  one  minor  modification  to avoid  any
spitback  problem.   However,  the submerged fill  system  needs  a
pressure  relief  system  like   that described  earlier.    The
submerged fill system is at  a  disadvantage in this respect.

     The  third characteristic of  comparison for  these  systems
is the complexity of  the  liquid seal itself.   The  in-tube trap
is the most complex   of the  three.  The submerged  fill  and 'J'
tube  designs  would be  much  less  difficult  to  build   and  are
probably more practical for  this reason.

     The  final basis of  comparison  is  the  security  of  the
liquid  seal  defined   in  terms  of  the   probability  of  vapor
escaping  to  the  atmosphere   during  refueling.   The  biggest
concern  about  trap  security   arises when  a   vehicle   sits  on
sloped terrain during  refueling.   The  in-tube trap and  'J' tube

-------
                              -35-
traps  can  be designed  to  ensure security up  to any reasonable
angle chosen.  The submerged  fill  system works differently than
these  systems,   however.   For  the  submerged  fill  to  function
properly,  the   fuel  must  enter  the  tank  below  the  liquid
surface,  but when the  tank is  nearly  empty,, the mouth  of  the
fillneck will be only slightly below the liquid  level.   If  the
vehicle were on an  incline,  the fillneck may  not be submerged
at all or may be only partially  submerged, and some  vapor could
escape during the early stages of refueling.

     It  is  clear from the discussion that each  system has both
strong  and  weak points, but  most  of the  problems with  any  of
the  systems  can be  resolved  with  a moderate  amount of effort.
The  relative complexity  of  the in-tube trap suggests  that this
design  may  be  the most  costly to produce,  but  if installed in
the  inside  top  of  the  tank  it could  be  implemented with  no
other  fillpipe  modifications.   The  other two systems  are also
clearly worth further  investigation  if  the  liquid seal concept
is being  developed.   Based on  the safety advantage,  the J-tube
is  somewhat preferable  to  the  submerged  fill,  and  EPA  has
selected  that  approach  for  inclusion  in the  prototype  liquid
seal onboard system  developed for  bench testing.   The  J-tube
was  also selected  because  of  EPA's previous experience with the
submerged fill  which showed  very  high  efficiencies  in  a bench
test.[4]

     The next  portion  of  this  report  covers EPA's  program to
evaluate  the   efficiency   of   a   J-tube  liquid   seal.    The
discussions  cover  the remainder of  the  components needed for a
bench  evaluation  of  the  system  and  the  efficiency  testing
conducted by EPA.

Ill.  Bench Testing of a Liquid Seal Onboard System

     A.    Description of the Prototype System

     1.    Introduction

     Once the evaluation of the three  liquid  seal concepts was
complete, EPA desired to  evaluate the  efficiency of  a  liquid
seal in a  simple  prototype  onboard refueling  control system.
As will be  described below,   the  prototype  system  constructed
for  this evaluation  was  relatively crude, and incorporated only
the  essential  components.   No   attempts   were  made  to  add
components or incorporate  modifications  to reduce the refueling
emission rate or optimize system  efficiency.   This  section  of
the  report   briefly   examines   the  three   components   added
downstream of the fuel tank to complete the prototype system.

-------
                               -36-
      To  complete   the   prototype   system   three • additional
 components   were   needed.   These  include   a   fill  limiter,   a
 properly sized vapor  line,  and  an activated  carbon  canister.
 Each of these components  is  discussed below,  in  the context of
 the flow rate and backpressure design characteristics mentioned
 above and  the overall  efficiency of  the prototype system.

      2.    Fill Limiter

      While  it does  not  directly affect  system backpressure or
 efficiency,  a fill  limiter  is  needed  in onboard system designs
 where the  vapor vent outlet  is  located   in  the fuel tank vapor
 drive.  On  present  fuel  tanks, the  automatic shut-off feature
.of   the fuel  nozzle   is   activated  when   tank  back  pressure
 increases  at the  end of the  refueling event and causes the  fuel
 in  the tank to back up  into  the fillneck  and  cover the nozzle
 spout with  liquid.   However, with  a  properly designed onboard
 system  fuel  tank  backpressures  are  minimal  since  vapor  is
 vented  during    refueling.    The  small   amount  of   system
 backpressure which does exist is not adequate  to  back  the  fuel
 up   into  the  fillneck  and  allow  operation   of  the  automatic
 shut-off feature  of  the  fuel  nozzle.   Therefore,  a  fill limiter
 is  needed  to close  the vapor vent  when the  tank  is  full  and
 thus  increase the backpressure.   Without a fill  limiter  it is
 quite possible that  the  tank  could be  filled well beyond  its
 nominal capacity.   For onboard  equipped  vehicles using   side
 fill   tanks  with   external vent  lines   (as  in Figure  12),   the
 onboard system has little  effect  on  the  refueling process.   The
 refueling  event  terminates the way  it does  on  present vehicles,
 and thus a  fill limiter  is not  needed.

      The  EPA  prototype  onboard system  used  a rear  fill   fuel
 tank without an external  vent  line, so  a simple float valve  was
 needed as  a fill limiter.   The  float  was constructed  of  the
 same  material used for  the tank's fuel  level  indicator.  It  was
 suspended  from the roof of the tank on  a simple  wire  in such  a
 way  that  it  would  float   on the fuel  as  the tank approached
 full,  and  would  nest  into  the vapor  vent when  the  tank  was
 full.    This  would   stop   the  venting   of  vapor   which  would
 immediately increase  system backpressure and thus would quickly
 lead  to activation of  the  nozzle automatic  shut-off.

      Before discussing  the  vapor  line,  two   points  about  the
 fill  limiter and  its  operation should  be  noted.   First, EPA's
 simple  prototype  did   not  include  a   liquid/vapor separator,
 rollover  valve   or   a   vent  closure  valve.   These  were   not
 necessary   for   a  simple  bench   set-up   intended  only   for
 evaluation  of  refueling  emissions  control.   Second, EPA did  not
 experience   significant  problems  with   fuel   spillage  due  to
 operation   of   the  fill  limiter,   as   has  been   suggested by

-------
                                          Figure 12
                            Typical  Current Evaporative  System
PRESSURE/VACUUM
   RELIEF CAP
             EXTERNAL VENT LINE
                      FLOAT/ROLLOVER
                          VALVE
                                               3/8" DIA.
                                               8' LONG
                                            .05" DIA. LIMITING
                                                 ORIFICE
                      14 GALLON FUEL TANK
                                                                          PURGE VALVE
                                                                    1 LITER
                                                                    CARBON
                                                                   CANISTER
TO PURGE
INDUCTION
POINT

-------
                              -38-
some.[5]   This perhaps  was  due  to  the fact  that the  system
evaluated by EPA included a reservoir in the fillneck.

     3.    Vapor Line

     The purpose of  the  vapor line  is to convey  gasoline  vapor
from  the  fuel  tank  to  the carbon  canister.   It must  be  sized
and constructed of material such that  it  does  not unnecessarily
increase the  system  backpressure  (which  affects the fill height
requirement),  and  must  be  impermeable  to gasoline vapors.   To
meet  these  requirements,   three  characteristics  of  the  vapor
line   and   its   use   must   be   considered   and   evaluated
concurrently.  These include  the  vapor line diameter,  the vapor
line configuration or layout,  and the  vapor  line material.   The
affect  of  these  three  characteristics  on  system  backpressure
and permeability of vapors is  discussed below.

     As gasoline vapors  flow  through the lines,  both major  and
minor  pressure  losses  occur  which  inhibit  vapor  flow  and
directly affect  system  backpressure.  The  vapor  line  diameter
makes  a major  contribution  to  the  pressure  losses,   since  it
represents a  major contraction in  the  flow.   The other  major
pressure  losses  are caused  by  functional  resistance to  the
vapor  flow.   This  is  a  function  of  the vapor  line  diameter,
length  of  vapor line used,  and to  some degree  the vapor  line
material.

     Minor pressure  losses  are  mainly caused by  the vapor  line
layout  or  configuration.   This  includes   any   bends,   elbows,
expansions or   contractions in  the  line,  as well  as any  other
deviations from the  straight  flow in a vapor line of a  constant
cross sectional area.

     As was  discussed earlier  in  relation  to  fillneck seals,
the  control   system  was  designed  so  that  the vapor  line  and
canister result in  a fuel  tank backpressure  of no more  than 5
inches  of  H2O.   A series  of  tests was  designed to empirically
determine  what size  vapor  line  would   be  appropriate  for  a
typical  light  duty  vehicle   application.    These  tests  were
performed  on complete  control systems   (J-tube  equipped  fuel
tank,   vapor   line,  carbon   canister)  subjected   to  actual
refuelings.   The  test  conditions  were  constant except   for  the
cross-sectional area  of  the  tubing  which was  varied from test
to  test.   Although  only  one  tubing  material  was tested,  the
tests  (summarized  in Table  3)  showed  that vapor  line  with an
inner diameter  of  at least 1/2  inch would  have to be  used to
keep  back  pressure  below  5  inches  H20  for the  entire system.
To be safe,  vapor  line  of 5/8 inch  inner diameter  was  used for

-------
                              -39-


                            Table 3


             Vapor Line Pressure Drop Contributions


                                          P,  Inches H2O
Tube ID,  Inch       Tube Configuration       Bench Test

      3/8

      1/2

      5/8

      1/2

      5/8
15
15
15
8
12
ft.,
ft. ,
ft. ,
ft. ,
ft. ,
6
6
6
4
4
bends
bends
bends
bends
bends
5
2
1
1


.2
.2
.5
.9

-------
                              -40-
control  system  testing,  although  1/2  inch vapor would  clearly
be reasonable for many vehicle applications.*

     The  tubing  material   is   also   very  important   to   the
functioning   of   the  control   system.    The  most   important
characteristic    of    appropriate    tubing     materials    is
impermeability  to  gasoline  and  gasoline  vapors.   If  a  vapor
line which was permeable to gasoline vapors was used  as  part of
the control  system,  hydrocarbons could escape  from  the  tubing,
defeating  the purpose  of   the   control  system.   There  are  a
number of  tubing  materials which would be  appropriate for  this
application.   Flexibility,  density  and cost characteristics  are
all  factors   to  be  weighed  in  deciding   between hard  rubber,
softer,  more  flexible  rubber  and  steel   materials.   A  double
walled,  flexible  rubber  fuel line  was  used in EPA1a prototype
onboard  control  system  since  5/8  inch  impermeable   vapor  line
was not commercially available.

     4.    Carbon Canister

     Activated  carbon  for  capture  and   storage  of  gasoline
vapors has  been used  successfully  in  automotive  applications
for more  than ten years.  To  apply this  technology to  control
of refueling  emissions in  EPA's  bench evaluation, the canister
must  not  increase system  backpressure  more than  is  necessary,
and it must be sized and loaded with enough activated carbon to
efficiently capture essentially  all  refueling  emissions.  These
two factors  are  closely linked  since  canister size  and design
both impact backpressure.

     Rather  than  designing  an  optimum canister,  it  was  EPA's
goal  to  construct or  obtain a  canister   which was  adequately
sized  to capture all  refueling  emissions  from an 18  gallon fuel
tank  and  yet  would  not  increase  total  system backpressure to
more than 5 inches H2O.

     First,  to  assist  in meeting  the  backpressure requirement,
EPA held  discussions  with  several  manufacturers of  activated
carbon.   Based  on  these  discussions,  EPA  selected  Westvaco
extruded wood base  activated carbon  for   use  in  the canister.
Information presented  by the  vendor indicated that the extruded
activated carbon  might  have  a  higher working  butane  working
capacity and  better  pressure  drop characteristics than the more
conventional  activated carbons now  used in evaporative emission
applications.  However,  it  appears  that   any  of  the activated
carbon types  sold  by  the  different companies  would  also  work
successfully   in capturing  refueling  vapors,  but   these  would
     This discussion briefly  addresses  how vapor  line diameter
     affects  backpressure.    Any  system  design  would have  to
     consider  vapor   line   length,   configuration,   canister
     effects, etc.,  in addition to vapor line diameter.

-------
                              -41-
have  different  performance characteristics  and  would require a
different amount of activated carbon.

      In sizing the canister, the next  step  would be to estimate
the  total  amount of activated  carbon  needed,  based on gasoline
vs. butane working capacity, apparent  density,  carbon aging and
other  factors.   As  is  shown on Table 4,  EPA estimates  that a
carbon  bed -of  1050  grams  would  be   required  to  capture   all
refueling  emissions  from  an  18 gallon  tank.   At  an apparent
density of 30  gm/100 ml,  a  canister  of  about  3500 milliliters
would be needed.

     Alternatively,  however,  EPA  turned  to  work  previously
conducted  by  Mobil  Research   and Development  in  their  1978
onboard  demonstration  program.   In  that  program Mobil  used a
4350 ml canister to  control  refueling  and evaporative emissions
from  a carburetted  1978  Pontiac   Sunbird which  had an  18.5
gallon  fuel  tank.[l]   An  inquiry to  Mobil  revealed  that  the
canister  shell  which is  shown  in Figure  13 was  still  on hand
and available  for  loan,  so this canister shell  was  used  by EPA
in  the   prototype   system.    This  canister  was  loaded  with
approximately  1400  grams  of  activated  carbon,   somewhat  more
than needed based on Table 4.

     5.    Conclusion

     We  have  now  described  all  portions  of   the  prototype
onboard system  used  to  evaluate the  efficiency  of  the  J-tube
liquid  seal.    The   prototype  onboard   system  included  the
following components:

     •     J-tube liquid fillneck seal with reservoir

     •     18 gallon fuel tank

     •     fill limiter

     •     5/8 inch vapor line

     •     4350 ml activated carbon canister

The  results  of  EPA's  evaluation testing  on  this  system are
described in the next portion of this  report.

     B.    Evaluation of System  Efficiency

     1.    Introduction

     Once the  prototype onboard system had been constructed in
a  bench  configuration,   a  series  of  refueling  emissions  tests
were  conducted  to  evaluate the  efficiency  of  the system and to

-------
                              -42-


                            Table 4


                   Total Required Carbon Bed
Refueling Emissions                           144 g
(at 8 g/gal for 18 gallons)
Activated Carbon Needed:                       700 g
50 Percent Carbon Aging Allowance and
Safety Factor                                 350

Total                                         1050 g
Based  on  a  gasoline  working  capacity  of  6.3  gm  gasoline
vapor/100  ml  carbon,  and   an   apparent  density  of   30  gm
carbon/100 ml carbon.

-------
                                        FIGURE 13
                          REFUELING SYSTEM CARBON CANISTER
                                               Carbon
                                                          /////// ////i/ / // / 77
                                                          W/////////77/777//////
                                                           Ill I|IMH	ITTTIPPf
                                                                                           Foam
                                               Foam
                                               Wire Mesh
                                        *—— Fiberglass Air Filter
Sunbird Carburetor
    Bowl  Fitting
                      Chevrolet  Carburetor
                           Bowl  Valves

                      0.047" Bleed to Fuel Tank
                  J_L
Side View of Top Disc
	    i!   i- 1/8" Plexiglass
ILL
                                     1"  Dia. xl 3/4" Long
                                       Plexiglass Tube
             Wire Mesh Glued to Bottom of Tube
                       of Tube
Chevrolet Purge Valve,
   Drilled to 0.180"
Canister Dimensions
    6" High
    8" Diameter
              Carbon: BPL-F3
                      1550 Grams
                      4350 ML

-------
                              -44-
draw  some  conclusions  as to  whether  liquid seal  systems  could
have  as  high an efficiency  as  mechanical seals.   This  portion
of  the  report describes how these tests  were  done,  the  results
of the testing, and draws some  conclusions  about  the efficiency
of liquid seal systems.

     2.     Description of Test Procedure

     The   refueling    tests   were  conducted   in   a   standard
evaporative emissions SHED which had been modified  to  allow the
entry of  a  fuel  hose fitting  into  the side  wall.  Fuel  was
dispensed from a standard fuel delivery cart.

     The  general  test   procedure   followed  is   outlined  in
Table 5, and  is  discussed below.  The  refueling  control system
bench system was placed in the SHED and  filled to  10 percent of
nominal   tank capacity.   The  fuel tank  was  then heated  to the
desired   temperature  inside  the  open  SHED  with  the  fuel  cap
on.*  The   purge  fan  was  operating  in  the  SHED  during  this
time.   The  dispensed   fuel  had  been  heated  to  its  desired
temperature previously.

     At   the  end  of   the  heating,  the  heating  blankets  were
unplugged,  the fuel  nozzle  was  inserted into the  fillneck,  the
mixing  fans  were  restarted,  and  the  SHED  was  sealed.   A
background  reading  was  taken  in   the  sealed   SHED   before
refueling  began.   The  refueling  was  then performed  by  turning
on the  fuel  cart  from  the outside the SHED.   The  refueling was
terminated  when  automatic  shut  off  occurred.   A  final  FID
reading   was  then  taken in  the SHED,  and  the  SHED floor  and
fillneck  area were  checked  for  fuel  spills.   The  refueling
emissions to  the SHED  were  evaluated  as  the  difference  between
the   initial   and  final   FID   readings,   with   appropriate
adjustments  for  the  small portion of  the SHED  volume  occupied
by the bench apparatus.

     Following  the  termination  of   the  testing,  the  carbon
canister was weighed and then purged with a total  of 300 cfm of
ambient   air  to  remove  refueling  vapors  from  the  canister.
Based on reduction in  canister  weights,  this amount of  air was
more  than  sufficient.   Following  completion  of the  purge, the
canister was weighed again.

     Finally, before describing  the  results  of the  testing, it
is important  to  discuss  the  temperature and RVP characteristics
     As  will  be  discussed  later,  a  couple  of  tests  were
     conducted with the  fuel  cap loose during the heat build to
     estimate the effects of heat build emissions.

-------
                              -45-



                            Table 5

                         Test Sequence

1.     Drain and refuel tank to  10 percent of nominal capacity.
2.     Connect heat blankets and thermocouples,  fuel cap on.
3.     Heat tank to desired temperature and stabilize.
4.     Insert fuel nozzle,  set latch at high level.
5.     Close SHED and start mixing fans.
6.     Take initial sample  reading (using FID).
7.     Remove cap, refuel tank to automatic shut-off.
8.     Check for spills and nozzle shut-off.
9.     Take final sample reading (using FID).
10.     Disconnect heat blankets  and thermocouples.

-------
                              -46-
of  the   test   fuels  used.    The  dispensed  and   fuel  tank
temperatures  were  set  at  92°F  +_  2°F  to  be consistent  with
previous  testing  conducted  by  EPA and to  be  representative of
the temperature proposed  by  EPA in its recommended practice for
testing of  refueling  emissions levels. [6, 7]   Fuels  of  11.8 and
9.0 psi RVP  were  used in the  program.   However  most tests were
conducted  using the 11.8  RVP  fuel  since  it  would  be expected to
create  higher  refueling  emissions  and   thus  would  be  a  more
stringent  evaluation of system efficiency.

     3 .    Refueling Emission Test Results

     The  results  of the  refueling  emission  tests  are summarized
in Table  6;  more  detailed information on each test is provided
in Appendix  A.   The  emission  results  shown include  7  tests at
11.8 RVP and 3 tests at 9.0 RVP.

     The  efficiency of  the  control  system  can be  evaluated by
comparing   the  refueling  losses  to  the  SHED  to  the  total
refueling  emissions based on  the  sum of  refueling losses to the
SHED  and  canister   weight   gain.   Under   this   approach  the
efficiency can be  calculated as shown below:

     EFF = 1 -          losses to  SHED _    x  100%
               caniscer win  gain T losses C^
     Using this  approach,  the average efficiency  is  98 percent
for the  11.8  psi  fuel  and 96.6 percent  for  the  9.0  psi  fuel.
Values  ranged  from  95.8  to  99.4  percent with  an average  of
about 97 percent.  However,  these  results clearly indicate that
theoretical efficiencies  of  98  percent  or  more  are  available
for  liquid seal   systems.   This  compares to  efficiencies  of
96-99 percent  for a mechanical seal system. [1,8]

     As was noted  previously,  two  tests  were  conducted with the
fuel  cap  loose during  the heat build  to assess  how  much heat
build emissions  were affecting  the total canister  vapor  load.
Referring  to  Table  6,  tests  1-5  were  conducted  with  the fuel
cap on  and fully  sealed during the  heat build,  while  tests  6
and 7  were conducted  with the  cap only loosely in  place.   A
comparison of  the  tests  shows  that the  canister weight gain and
the apparent refueling  emission  rate were greater when the cap
was on  and sealed.  To  some  degree the  canister  was  accepting
vapor from  the tank  during the  heat build similar  to  what now
occurs  in  a diurnal evaporative  emissions  test.   However,  as
can be  seen from Table  6,  this has  little effect  on the  system
efficiency evaluation for  the program.    All  values  fall  within
the 96-99 percent efficiency range.

-------
RVP

11.8 psi
                              -47-



                                  Table 6

                   Liquid  Seal  Control  System Efficiencies
Conditions
Test
1
2
3
4
5
6
7
TT
91.0
90.8
92.0
91.0
91.9
92.0
91.0
ID
92.0
90.3
90.5
91.7
91.9
92.0
92.5
Gal
12.5
15.5
15.0
15.1
14.0
15.5
12.5
Loss to
SHED(g)
2.3
3.6
6.1
4.9
2.5
2.0
1.5
Canister
Wt Gain(g)
153.3
168.1
164.1
129.3
123.9
170.5
163.0
Percent
Efficiency
99.4
97.9
96.4
96.3
98.0
98.8
99.1
                                                      Avg. EFF      98.0

 9.0 psi    08     92.0    93.0   18.5     6.0        135.5         95.8
            09     93.0    92.2   15.0     3.6        121.1         97.1
            10     92.0    92.5   15.0     4.0        120.8         96.8

                                                      Avg. EFF      96.6

-------
                              -48-
     The turbulent mixing  in the fillpipe  discussed previously
and  the   subsequent   entrainment   of   bubbles  in  the  liquid
entering  the  tank  suggests  that  the   liquid   seal   may  be
generating  additional  refueling   emissions   relative   to  the
uncontrolled  fuel  tank.   Uncontrolled  refueling  emissions were
not evaluated on  the  bench system,  but a  similar  fuel  tank was
evaluated  on  the  1983  Oldsmobile   Cutlass   evaluated  in  a
previous  EPA  test program.[7]   The  results  of  that  testing
indicate that uncontrolled  refueling emissions  for this vehicle
can be predicted using the equation

Losses  (g/gal)  =  -5.584  - 0.114[AT(°F)]   +  0.0857  [TD(°F>]  +
0.520 [RVP(psi)],  where AT = TT   - TD.

     Substituting the temperature and RVP  conditions of Table 6
into  this  equation  (TD  =  92°F,  AT  = OF0,  RVP  = 11.8  psi)
yields an emission rate of 8.4 g/gal.

     For comparison, two  tests  were conducted  on  the prototype
onboard  system without  the canister  in  place.   The refueling
emission  rate averaged  9.9  g/gal   for the   two  tests  (Tests
85-1363, 1364 in Appendix A).

     The  comparison  of   the  actually  occurring   and predicted
emissions discussed above, leads to  the conclusion that in this
case  the  liquid  seal  increased the  emission rate.   However,
using this  data  it is  not possible to  precisely  quantify the
effect;   the  use of  an average  could  be  misleading  due  to the
relatively small number of  tests and the  range in values seen.
Ideally,  comparisons   should be made  between identical  tanks
with and without  a liquid seal  with as many  other variables as
possible controlled.

     A  close  review   of   the  detailed  results  in  Appendix  A
reveals  that  14-15.5  gallons was  dispensed in most tests, but
over 18  gallons was  dispensed in one test.  This  later test was
the first  conducted  in  the  series and led  to  the conclusion
that  the  fill limiter   installed   in  the  tank  did  not  work
properly  and  an  overfill  occurred.   However,  when  the  fill
limiter  was  modified  and reinstalled  it  dropped  too  low into
the tank and  about a  gallon of   nominal tank  capacity was  lost.
This accounts  for  the  difference in the dispensed gallons  among
the tests,  and explains  why the  amount  of  fuel  dispensed in
most tests  was somewhat  less than 90 percent of nominal tank
capacity.

     4 .     Source of Emissions

     In  addition  to  assessing   the  control efficiency  of the
system,   tests were  also  conducted to  locate  the source  of
refueling emissions from the control system.   The  first step in
this process  involved  checking  the  fuel tank and  vapor line for

-------
                              -49-
leaks.   This was  done  by  removing  the canister from the system,
plugging the vapor  line  (at  the canister end)  and  pressurizing
the tank.  A  liquid was  then applied to the tank and vapor line
so that leaks could be seen.   The system was  free from  leaks  so
the only, possible  sources  of  emissions  were the  fillneck  and
the canister  (if  breakthrough  were  occurring).   In  order  to
isolate the  emissions  from the fillneck, the  canister  was left
off of  the  system  and was  replaced with a  plastic vapor  bag.
Refueling operations were performed  with the vapor  bag  in place
and the  refueling  loss  was   measured.   In   this  way,  only  the
emissions from  the  fillneck  were  measured.   The  results  from
this set of  tests are shown in Table 7.

     The  refueling   losses  from the  control  system  with  the
vapor  bag  are very similar   to  those  emitted from  the  control
system with the  canister  in  place.   The  projected  efficiencies
shown  in Table   7   were  developed  by  dividing the  refueling
losses  by the predicted emission level  found using  the  equation
cited previously.

     Since the refueling  loss  and  efficiency numbers of Table 7
are very similar  to those in Table  6,  it  appears that  most  of
the refueling loss comes  from the fillneck.

IV.  Conclusions

     The purpose  of  the development  and test programs conducted
was  to evaluate  the  practicality   and feasibility  of  liquid
fillneck  seals   as   an  alternative  to   the  more   widely
demonstrated  mechanical   seals.   The  EPA   program  covered  a
laboratory evaluation  of  three different liquid seal  concepts
and bench testing of a simple prototype onboard system  using a
liquid  seal.

     The evaluation  of the liquid  seal concept led  to  several
conclusions.

     •      The   available  fill  height  on  most    side   fill
           passenger  cars and  light  trucks  is sufficient  to
           permit  the  use  of  liquid   seals  without  premature
           nozzle shut-off.

     •      For  some  side fill vehicles,   reductions   in  fill
           height would  need to  be  achieved through   internal
           fillpipe modifications to control the bubbles caused
           by the strong  turbulent  mixing in the fillpipe.

     •      Rear   fill  vehicles  would  require more  substantial
           fillpipe, tank, or other  modifications  to use liquid
           seals   and thus may  be   better  suited to  mechanical
           seals.

-------
                         -50-
                         Table  7



  Sources of Refueling Emissions on a Controlled System*

Test
11
12

TT
91.0
91.0
Condi
TD
92.5
92.5
tions
Gal.
13.3
13.5
Loss to
SHED(g)
2.4
2.8
Predicted
Emissions (g)
115.0
116.8
Percent
Efficiency
97.9
97.6
Vapor bag used to replace canister; RVP = 11.8.

-------
                              -51-
     •     Liquid  seal  systems   (especially   the   J-tube  and
           in-tube trap) appear .to have safety  advantages over
           the mechanical seal and submerged fill.

     •     The strong turbulent mixing  in  the  fillpipe entrains
           air into  the fuel entering  the  tank, and  increases
           the volume of vapor which  is displaced from the tank
           as it is refilled.

Several  key  conclusions can  also  be  drawn   from  the  bench
testing of a prototype onboard system using a liquid seal.

     •     The  efficiency   of   the  liquid   seal   systems  is
           comparable  to that  achieved  by  mechanical  seals.
           The mechanical seal  systems evaluated by API showed
           efficiencies  of  96  to   99  percent.   The   liquid seal
           systems    evaluated     in    this    project    showed
           efficiencies in that range.

     •     With    improvements    in    the    control    system,
           consistently  high control  efficiencies   (98  percent
           and greater)  are  easily within  reach for liquid seal
           systems.

     •     For a  liquid seal  system with  an   adequately  sized
           carbon  canister,   most   refueling   losses  appear  to
           arise from the fillneck.

     •     The air  entrainment caused  by the  turbulent mixing
           in  the  fillneck  appears   to   increase  the  total
           refueling  emissions  load  to  the  carbon  canister.
           Data generated in  this  test program is  insufficient
           to precisely quantify this effect.

     The primary purpose of  this  program was to demonstrate the
feasibility  of  liquid  fillneck   seal   concepts.    The  bench
prototype developed here was  clearly  a first generation system;
no  attempts  were  made  to  optimize  the  design  to  reduce  the
effects  of  entrainment or  to maximize  control  efficiencies.
Improvements  in  both  areas  are  possible,  as  demonstrated  in
later work  conducted  by API. [8]   This  work shows  that  onboard
systems  using  liquid fillneck  seals  can  consistently  control
refueling emissions with efficiencies of 98 to  99 percent.

     The  development   and   test    programs  conducted  by  EPA
demonstrate  that  liquid  seal  systems  are  indeed  practical
alternatives  to   mechanical   seal   systems.   Analysis  suggests
that liquid seals are superior to  mechanical seals  in the  areas
of   durability,    tampering,   and   safety,   and   have   equal
efficiencies for controlling  refueling  emissions.   However, the

-------
                              -52-
work conducted by  EPA indicates that  liquid  seals may  be  more
attractive for  side  fill than  rear  fill vehicles,  and  the air
entrainment  caused by  turbulent  mixing  in  the  fillpipe  may
actually increase  refueling emissions  and create the need for a
larger canister than  that  required for mechanical  seals.   Thus
some  trade-offs   may  exist   between  mechanical   and   liquid
fillneck seals.

     Either  liquid or mechanical  fillneck  seals could  be  used
as part  of  an onboard system.  Each approach  has  its strengths
a-nd weaknesses, but  both have  been demonstrated  to  have  high
efficiencies  on   prototype   systems.    Whether   a   liquid  or
mechanical seal system is preferable would  have  to  be evaluated
separately for each vehicle model.

-------
                              -53-
                           References
     1.    "On-Board    Control    of    Refueling    Emissions,
Demonstration of  Feasibility,"  Mobile Research  and  Development
Corp.,  for  American  Petroleum  Institute,   Washington,   D.C.,
October 2 1978.

     2.    "Control  of  Refueling  Emissions   with an  Activated
Carbon   Canister   on  the   Vehicle   -  Performance   and   Cost
Effectiveness Analysis," Olson  Laboratories,  Inc.,  for American
Petroleum Institute, Washington, D.C., October,  1973.

     3.    "SAE Recommended  Practice  (J  1045)  - Instrumentation
and Techniques for Vehicle Refueling  Emissions Measurement."

     4.    "Status  of  In-House Refueling  Loss  Measurements,"
Memorandum  from  Martin Reinemann  to Robert  Maxwell,  U.S.  EPA,
SDSB, March 6, 1979.

     5.    "Ford  Motor  Company Comments  on   the Evaluation  of
Air  Pollution Regulatory  Strategies   for  the  Gasoline Marketing
Industry," Nov.  1984.

     6.    "Draft    Recommended    Test   Procedure    for    the
Measurement of Refueling Emissions,"   Lisa Snapp, EPA Technical
Report, EPA-AA-SDSB-85-5,  July, 1985.

     7.    "Refueling Emissions from  Uncontrolled Vehicles,"  D.
Rothman and R. Johnson,  EPA Report EPA-AA-SDSB-85-6,  1985.

     8.    "Vehicle  Onboard  Refueling Control," API  Publication
4424, March 1986.

-------
        -54-
     APPENDIX A
Detailed Test Results

-------
                                               Fuel Tank Bench Tests
                                 FUEL-COMMERCIAL UNLEADED, SUMMER, 92°F DISPENSED*
                       Temperatures °F
Dis- Refuel-
pensed ing

Test No.
85-0694
85-0120
85-0121
85-0123
85-0698
85-0699
85-1358
85-1362
85-1368
84-5633
84-5634
84-5635
85-1363
85-1364

Date
10-31-84
11-05-84
11-06-84
11-07-84
11-08-84
11-09-84
12-19-84
01-04-85
01-11-85
08-23-84
08-24-84
08-24-84
01-04-85
01-03-85
Ini-
tial
66.8
68.2
68.3
69.5
67.55
67.0
61.0
67.7
60.5
69.0
67.0
67.9
70.7
67.7

Final
91.0
91.0
90.8
91.0
92.0
91.0
92.0
91.0
91.9
92.0
93.0
92.0
92.0
92.5
Final
Vap
89.0
88.8
89.0
89.0
90.0
89.5
88.5
89.5
91.0
90.2
89.0
90.3
89.0
90.8
Max
T
3.0
3.0
3.0
3.5
3.0
3.0
3.5
3.0
5.3
2.0
2.0
2.0
4.0
3.0
Dis-
pensed
92.0
92.5
90.3
91.0
92.0
92.5
90.5
91.7
91.9
93.0
92.2
92.5
89.8
90.0
Gal-
lons
12.5
13.3
15.5
13.5
15.5
12.5
15.0
15.1
14.0
18.5
15.0
15.0
14.3
14.5
Loss,
gms.
2.3
2.4
3.6
2.8
2.0
1.5
6.1
4.9
2.5
6.0
3.6
4.0
134.8
151.5
Canister
Wt Gain
153.3
NA
168.1
NA
170.5
163.0
164.1
129.3
123.9
135.5
121.1
120.8
NA
NA
Dis-
pensing
Time
Mi n, Sec
1'46"
2'04"
2'11"
1'57"
2'19"
1'49"
2'16"
2'08"
2'50"
3'41"
2'56"
2'55"
2'00"
2'09"
Heat
Time
Hr,Min
57"
52"
56"
45"
51"
55"
58"
40"
45"
46"
48"
48"
56"
42"
                                                                                           Spi llage/Coimients

                                                                                           None
                                                                                           None, vapor bag
                                                                                           None, manual shutoff
                                                                                           None, vapor bag
                                                                                           None, can  in box
                                                                                            multi-fueling
                                                                                           None, multi fueling
                                                                                           None, Benzene 13A
                                                                                           Controlled
                                                                                           None, canister only
                                                                                            during  refueling
                                                                                           None
                                                                                           None, manual shutoff
                                                                                           None, manual shutoff
                                                                                           Controlled w/o canister
                                                                                           Controlled w/o canister
All tests were at 11.8 RVP except 84-5633, 5634, and 5635.

-------