Simulation of Peak Stresses and Bowing Phenomena

during the Cool Down of a Cryogenic Transfer System

Hubertus Tummescheit

, Kristian Tuszynski

, Philipp Arnold

Modelon AB
,
Ideon cience Park, !"##$%& 'und, (eden
'inde Kryotechik A)

, *H"+,## P-un.en, (itzerland

Hubertus/Tummescheit0modelon/se, Kristian/Tuszynski0modelon/se
Abstract
An e1tension -or cryo.enic systems to the Air*ondi"
tionin. 'ibrary by Modelon (as used to analyze the
cool do(n o- a cryo.enic trans-er system (here
'inde Kryotechnolo.ie in P-un.en, (itzerland (as
the main contractor/ imulation (as used early in the
desi.n process to make sure that the system (as (ell
desi.ned -or a number o- cool"do(n scenarios/ !arly
detection o- problematic parts o- the system -or some
cool"do(n se2uences lead to chan.es in the pipin.
desi.n/ imulation (as also used to assess the ma1i"
mum thermal stresses durin. cool do(n and deter"
mine suitable mass -lo( rates/ Proper cool"do(n se"
2uences (ere established iterati3ely (ith the help o-
a combined simulation o- the cryo.enic t(o"phase
-lo(, the heat conduction in solid structures and the
resultin. thermal stresses/ The t(o main problems to
a3oid durin. cool do(n are 456 e1cessi3e thermal
stresses in thick"(alled components, and 4#6 bo(in.
o- pipes (ith li2uid cryo.en in the lo(er part o- a
lon., horizontal pipe (ith .aseous cryo.en abo3e/
T(o similar systems (here considered, one -or li2"
uid hydro.en, the other -or li2uid o1y.en/ 7ymola
and Modelica (ere choosen -or the pro8ect due to the
.ood multi"domain and multi"physics capabilities,
and the a3ailability o- model libraries that co3ered a
lar.e part o- the problem/
Keywords: Cryogenics, two-phase flow, transient
thermal stress simulation
1 Introduction
The Indian pace 9esearch :r.anization, I9:, is
buildin. and commissionin. a ne( cryo.enic en.ine
test ri. in their 'i2uid Propulsion Test *entre in Ma"
hendra.iri, Tamil ;adu/ The system under in3esti.a"
tion is the cryo.enic trans-er system -or the cryo"
.enic -luids hydro.en and o1y.en, used to trans-er
cryo.en -rom tankers into the run"tanks and -rom
both tankers and run tanks to the test ob8ects/ The
system is desi.ned -or a (ide ran.e o- pressures and
-lo( rates (hich leads to a rather comple1 o3erall
structure o- pipes, 3al3es and measurement e2uip"
ment/ imulations o- the system cool do(n (as used
early in the desi.n process to 3alidate the desi.n <
here the main issue is to a3oid bo(in. o- dead"end
pipes < and to -ind impro3ement potential -rom an
operational point o- 3ie(/ imulation (as also used
later on to establish suitable cool"do(n -lo( rates
and 3al3e se2uences that -ul-ill the t(o main re2uire"
ments= use as little cryo.en as possible -or cool do(n
(hile not e1ceedin. the ma1imum allo(ed thermal
stresses/
:btainin. the desired mass -lo( rates in a transient
t(o"phase -lo( system throu.hout the system is 3ery
di--icult because o- the enormous chan.e in densities
bet(een .aseous and li2uid cryo.en= the density ra"
tio can be up to 5=5&&&/ 7urin. the -illin. o- the sys"
tem (ith li2uid, de3iations bet(een local mass -lo(
rates and controlled rates at a 3al3e (ith one"phase
inlet conditions can be lar.e/ In the situations (hen
the control 3al3e is inside the t(o"phase re.ion, ac"
tual mass -lo( rates can not be controlled at all/
2 odeling of thermal stress in
cylindrical bodies
The model -or thermal stress is based on a radial dis"
cretization o- cylindrical .eometries both -or pipes
and 3al3es/ >or the bo(in. phenomenon, also a tan"
.ential discretization and, i- necessary an a1ial one
are added/ The ener.y balance o- a cylindrical slice
o- the pipe is based on the >ourier e2uation (ith a
central di--erence appro1imation o- the temperature
.radient and takes the temperature dependence o- the
heat capacity and thermal conducti3ity into account/
tresses are computed separately -or the stress intro"
duced throu.h temperature .radients and the me"
chanical stress due to the pressure inside the pipe/
The stress 3ectors are summed to compute a total
e2ui3alent stress/ The e2ui3alent stress reaches its
Simulation of Peak Stresses and Bowing Phenomena during the Cool Down of a Cryogenic Transfer System
The Modelica Association 693 Modelica 2008, March 3
rd
4
th
, 2008
ma1imum 3alue either on the inside o- the cylinder
or on the outside o- the cylinder/ The ratio o- the
ma1imum e2ui3alent stress and the yield stress is the
stress ratio/
The >ourier e2uation is .i3en by ?%@,
j
i
\
i
(
T
i
t
Cp
i
+
Cp
i
t
T
i
)
= A
i
T
i5
+B
i
T
i
+C
i
T
i+5
456
Ahere i = 2, 3 !r-" and !r is the discretization
number o- the material in radial direction/ The t(o
remainin. elements are .i3en by the boundary condi"
tions/
The calculation o- the >ourier coe--icients, A, B and
C -or a radial discretization is sho(n in e2uation #2$/
A
i
=
r
i
+r
i5
r
i
( r
i
r
i 5
)(r
i +5
r
i 5
)
B
i
=
r
i +5
+r
i
r
i
( r
i+5
r
i
)(r
i +5
r
i 5
)
C
i
=A
i
B
i
4#6
The a1ial heat conduction in the material is assumed
to be ne.li.ible/
To obtain the thermal stress distribution, three stress
components in tan.ential 4B6, radial 4r6 and a1ial 4z6
directions are calculated/ The .eneral stress e2ua"
tions are .i3en by

c
0
=
%o
(5+)r
#

|
r
#
+r
i
#
r
o
#
+r
i
#

r
i
r
o
T (r) r dr+

r
i
r
T (r) r drT ( r)r
#

c
r
=
%o
(5+) r
#

|
r
#
r
i
#
r
o
#
+r
i
#

r
i
r
o
T (r) r dr

r
i
r
T (r) r dr

c
&
=
%o
(5+)

|
#
r
o
#
+r
i
#

r
i
r
o
T (r) r drT (r )

4$6
(here % is the Coun. modulus, ' the linear e1pan"
sion coe--icient and ( the Poisson ratio/
By only calculatin. the thermal stress at the inner
and outer points o- the (all 4the ma1imum stress o- a
pipe is al(ays at one o- these points6 the e2uations
can be simpli-ied as=
c
0
i
=
%
5
o
5
(5+
5
)

|
TmT ( r
i
)

c
0
o
=
%
! r
o
! r
(5+
!r
)

|
TmT (r
o
)

c
&
i
=
%
5
o
5
(5+
5
)

|
TmT ( r
i
)

c
&
o
=
%
! r
o
! r
(5+
! r
)

|
TmT (r
o
)

c
r
i
=&, c
r
o
=&
4,6
Ahere Tm is the mean temperature o- the material
and 5 and ;r re-er to the innermost and outermost ra"
dial discretizations respecti3ely/
The e--ecti3e stress accordin. to Don"Mises theory
results in 4-rom ?%@6=
c
eff
=
.
c
0
#
+c
&
#
+c
r
#
(c
0
c
r
+c
0
c
&
+c
r
c
&
)
4E6
The stress"ratio is de-ined as the ratio o- e--ecti3e
stress to yield stress o- the material=
c
ratio
=
c
eff
c
)
4F6
The t(o di--erent problems analyzed later in the pa"
per need di--erent discretizations/
5/ The thermal stress analysis -rom cool"do(n
re2uires a t(o"dimensional model (ith radi"
al and a1ial discretizations to capture the lo"
cal thermal stresses alon. the pipe/
#/ The bo(in. problem re2uires a1ial and tan"
.ential discretizations to capture the di--er"
ent de-ormations on the top and bottom o- a
pipe (here the bottom is -illed (ith boilin.
li2uid and the top is -illed (ith saturated .as/
Both cases (ere captured (ith a sin.le model (ith
all $ discretizations, (here the ones that (ere not
needed (ere set to one element/
H. Tummescheit, K. Tuszynski, P. Arnold
The Modelica Association 694 Modelica 2008, March 3
rd
4
th
, 2008
! "low modeling
>or the t(o"phase -lo( in the pipes, a standard -inite
3olume method assumin. homo.eneous e2uilibrium
-lo( (as used as described in ?#@ and ?$@/ 7ue to the
partially 3iolent transients, a dynamic momentum
balance has been used -or some o- the simulations/
Heat trans-er needs to take into account the Gsub"
cooled boilin.H re.ime, (hich is important to(ards
the end o- the cool do(n and is present durin. a
lar.e -raction o- the o3erall cooldo(n time/ Pressure
drop models are -rom the standard literature like ?,@/
Properties -or o1y.en (ere implemented accordin.
to ?5@, hydro.en properties accordin. to ?+@, and the
results (ere compared to 9e-Prop by ;IT (hich
contains the same property models/
The main trade"o-- that has to be taken into account
is bet(een minimal cryo.en consumption -or cool"
do(n and a minimal cool do(n time/ The mass -lo(
is restricted by an upper limit, usually determined by
the ma1imum allo(able thermal stress, and a lo(er
limit/ The lo(er limit is de-ined by the Gnon"strati"
-ied -lo(H condition/ A stable phase separation (ith
li2uid -lo( on the bottom o- the pipe and .aseous
-lo( abo3e it results in di--erences in the heat trans"
-er rate o- about one order o- ma.nitude/ They may
lead to -aster cooldo(n on the bottom o- the pipe,
(hich may lead to bo(in./ The limit -or strati-ied
-lo( conditions -or cryo.ens has been in3esti.ated in
?I@/
It could be ar.ued that a homo.eneous e2uilibrium
model does not capture the physics o- the cool do(n
-lo( su--iciently accurate/ >or the main -ocus o- the
study, the thermal stress in the thick"(alled compo"
nents, it is not necessary to predict the -lo( and the
-lo("re.ime e1actly 4appart -rom a3oidin. strati-ied
-lo( conditions6, and there-ore (e do not belie3e
that a non"homo.eneous -lo( (ould impro3e the
2uality o- the results in a (ay that (ould 8usti-y the
much hi.her model comple1ity/
# $ow tem%erature %ro%erties
Both the thermal conducti3ity and the heat capacity
o- metal pipes .o to & at & de.rees Kel3in/ This has a
number o- surprisin. e--ects (hen the temperatures
are approachin. the lo(er limits 4ca/ #& K -or li2uid
hydro.en and ca/ +& K -or li2uid o1y.en6= the cold
parts o- metal pipes and 3al3es almost insulate the
remainin. (armer parts -rom the cryo.en, e--ecti3e"
ly slo(in. do(n the last part o- the cool do(n/
>ortunately detailed data -or metals used in cryo"
.enic trans-er systems is publicly a3ailable -rom
;IT 4;ational Institute -or tandards and Technol"
o.y6 3ia their (eb"based database, see ?+@/
*igure ": *low rate which predicts non-strati-
fied flow conditions for pipeline fluid +ualities
,elow-./ #li+uid and gas phase assumed saturated
at ,oiling point$, from 0-1
*igure 2: Thermal conducti2ity for steel 3"34
as a function of temperature
Simulation of Peak Stresses and Bowing Phenomena during the Cool Down of a Cryogenic Transfer System
The Modelica Association 695 Modelica 2008, March 3
rd
4
th
, 2008
& Thermal stress results
>or the e3aluation o- thermal stresses, a 5& m lon.
pipe (ith ca 5E& mm diameter 47; 5E&6 and ca +
mm (all thickness, material stainless steel $5F -or
o1y.en and $&, -or hydro.en, directly do(nstream
o- an open"close 3al3e that opens completely in #
seconds is in3esti.ated/ The results -or pipes .i3e a
.ood understandin. -or the situation o- the complete
system as they demonstrate (ell the di--erences
caused by the di--erent cryo.en properties The up"
stream properties are=
5/ 'i2uid o1y.en o- &/E MPa at I5 K/
#/ 'i2uid hydro.en o- &/E MPa at #5 K/
In both cases, the hi.hest stress is not directly do(n
$/ stream o- the 3al3e but a short distance into the pipe,
at a location (here the combination o- hi.h heat
trans-er coe--icient and lar.e T results in the com"
bination (ith the hi.hest heat -lo(/ The lon.itudinal
discretization is #& se.ments, the radial discretiza"
tion 5& se.ments -or pipes/
7i--erent phases o- cool do(n can be clearly dis"
tin.uished -rom the temperature tra8ectories/ The di-"
-erence bet(een hydro.en and o1y.en cool do(n is
also strikin., but becomes understandable once the
in-luences o- the di--erent thermophysical properties
o- the -luids and the metals are taken into considera"
tion/ ome o- the results are not entirely intuiti3e,
e/./ that the -irst part o- the pipe has initially lo(er
temperature than the do(nstream parts, but is the last
part to be cooled do(n entirely/
This is e1plainable -rom the chan.e o- the heat trans"
-er coe--icients o3er timeJtemperature= ob3iously the
-irst part o- the pipe is cooled do(n -aster at the be"
.innin., but the combination o- a cold (all 4metal
conducti3ity decreases (ith temperature6 and a lo(
heat trans-er coe--icient 4the be.innin. o- the pipe is
e1posed to sin.le phase li2uid -lo( at 3ery lo(
9eynolds numbers6/ This -act, combined (ith -urther
results omitted here, leads to the result that the cool
do(n time is independent o- the pipe len.th -or pipe
len.ths less than E& m/ The di--erences bet(een o1y"
.en and hydro.en cool do(n become clear (hen
lookin. in more detail at the re2uired ener.y -or the
metal cool do(n and the a3ailable speci-ic enthalpy
di--erences -or coolin. in di--erent phases, tabulated
in Table 5/ It is ob3ious that a much lar.er part o- the
cooldo(n is bet(een .as phase and metal -or hydro"
.en, both due to the lar.er ener.y content and the
lar.er temperature di--erence/ The .as phase cool
do(n has a lo(er heat trans-er coe--icient (hich
*igure 3: 5pecific heat capacity for steel 3"34
as a function of temperature
*igure .: Temperature o2er time with different
phases for a "6m low-pressure pipe during cool
down with a hydrogen mass flow of 62. 7g8s at
control 2al2e
*igure 9: Temperature o2er time with different
phases for a "6m high-pressure pipe during
cool down with an o:ygen mass flow of "; 7g8s
at control 2al2e
H. Tummescheit, K. Tuszynski, P. Arnold
The Modelica Association 696 Modelica 2008, March 3
rd
4
th
, 2008
leads to lo(er stress peaks in the material/ econdly,
the rapid cool"do(n (ith t(o"phase -lo( mostly
happens a-ter e--icient pre"coolin. (ith cold .as/
:3erall and a.ainst -irst intuition, coolin. do(n (ith
o1y.en poses hi.her risks in spite o- the lo(er tem"
perature di--erence/ ;ote also the temperature
Gbounce"backH o- the metal layer in contact (ith the
hydro.en a-ter the hydro.en in the pipe chan.es
-rom t(o"phase to li2uid/ This e--ect is caused by the
drastic drop in heat trans-er coe--icent in the pres"
ence o- much (armer outer layers in the pipe metal/
aterial '
%hase
(nergy content for com%lete
cooldown )from !1* + to *, + for
-
2
. 2,+ for /
2
0
teel $&, 101.5 kJ/kg
teel $5F 94.6 kJ/kg
Total h H
#
4158 kJ/kg
Total h :
#
391.2 kJ/kg
h H
# e3ap
$%$ kKJk.
h H
# .as
3785 kKJk.
h :
# e3ap
5I5/# kKJk.
h :
# .as
#&& kKJk.
Ta,le ": <ntegrated energy content comparison
The lar.est source o- uncertainty in the e3aluation o-
the stress ratio is the occurrence o- the Gboilin. cri"
sisH in t(o"phase heat trans-er at 3ery hi.h heat -lo(
rates/ Lnder such conditions, a thin layer o- .as at
the metal (all separates the boilin. li2uid -rom the
metal by an insulatin. layer, thus drastically reduc"
in. the heat -lo( and the resultin. thermal stresses/
*orrelations -or the occurrence o- the boilin. crisis
-or cryo.enic -luids are not 3ery reliable, data only
e1ists -or non"cryo.enic -luids/ In addition, the boil"
in. crisis condition -or cryo.enic coolin. occurs at
4almost6 constant temperature o- the hot side, (hich
is di--erent -orm the usual e1periments (ith rapid
heatin. and risin. temperature on the hot side/
Ahile this means that the e1act heat trans-er in
the 3icinity o- the boilin. crisis is di--icult, the
e1istin. correlations can nonetheless be used to
estimate the hi.hest reasonable heat trans-er co"
e--icient and thus the (orst case scenario -or the
thermal stresses in the metal (all/ The results in
>i.ure F -or a hi.h pressure pipe sho( that the
combination o- (orst case assumptions 4-irst
se.ments o- pipe that is sub8ect to t(o"phase
heat trans-er -rom the start and hi.h coe--icinent
o- heat trans-er6 lead to stress ratios close to the
permissible limit/ The stress ratio plot in >i.ure
F also sho(s that locations -urther do(nstream
are sub8ect to lo(er stress due to pre"coolin.
(ith cold .as/ The stress peaks (iden and the
le3el decreases as the t(o"phase zone (idens
-urther do(nstream/ A sensiti3ity study (as
conducted (ith repect to the most important pa"
rameters -or the stress calculation, amon. others,
the heat trans-er coe--icient, and the result (as
that the ma1imum heat trans-er coe--icient had a
ne.li.ible e--ect on the stress ratio/ >or 3al3es,
due to the much thicker metal (alls, the stress
ratio e1ceeds 5/& locally and -or brie- times/
*ryo.enic 3al3es sur3i3e these conditions, but
the hi.h thermal stress leads to local de-orma"
tions and Gcold hardenin.H, but is -ar -rom 3al"
ues that (ould cause complete material -ailure/
Ahile it is not possible to a3oid these conditions
e3ery(here in the system, the operation o- the
plant can be adapted to minimize the number o-
times and locations that are sub8ect to the e1"
treme conditions/ It (as, ho(e3er, possible to
a3oid the se3ere thermal stress conditions -or
3al3es in the hi.h pressure part o- the system/
*igure 3: 5tress ration along o:ygen pipe di-
rectly after 2al2e without any pre-cooling
Simulation of Peak Stresses and Bowing Phenomena during the Cool Down of a Cryogenic Transfer System
The Modelica Association 697 Modelica 2008, March 3
rd
4
th
, 2008
1 Pi%e Bowing
The calculation o- the pipe bendin. due to the tem"
perature di--erence at the top and bottom o- the pipe,
(hen -illin. (ith cold li2uid, is done (ith the -ol"
lo(in. assumptions=
Pipes are considered strai.ht,
The pipes are -i1ed at the lo(er end points
(ith a .lidin. support at one end to compen"
sate -or lon.itudinal len.th chan.e,
Both radial and circum-erential heat trans-er
is taken into consideration in the (all, a1ial
heat trans-er is ne.lected due to a1ial sym"
metry/
The li2uid cross section area in the pipe is calculated
accordin. to #"$=
A
li+
=acos(5
4
li+
r
pipe
)r
pipe
#
+( 4
li+
r
pipe
)

.
#r
pipe
4
li+
4
li+
#
456
The li2uid 3olume is computed -rom the mass -lo(
into the horizontal pipe, assumed to end at a closed
3al3e/ The mass -lo( into the pipe is taken -rom a
prior cooldo(n simulation, at the position o- the hor"
izontally connected pipe/
imulations use t(o heat trans-er coe--icients, one
-or the part o- the (all that is in contact (ith li2uid
and the other -or the part that is in contact (ith the
.as/ The .as temperature has 3ery little in-luence on
the o3erall result because o- the lo( heat trans-er co"
e--icient bet(een .as and pipe (all/ 7ue to the boil"
in. li2uid underneath it (ill (ithin short time a-ter
li2uid cryo.en is at the bottom reach the saturation
temperature/
To calculate the heat trans-er to the (all the actual
li2uid le3el is used to -ind the len.th o- a discretiza"
tion that is co3ered by li2uid 4i- any6 and the heat
trans-er is proportional to this 3alue=

==7
li+
(T
wall
T
li+
)A
>T

4
li+
4
?,ottom
4
?top
4
?,ottom
+7
gas
(T
wall
T
gas
)A
>T
(5
4
li+
4
?,ottom
4
?top
4
?,ottom
)
(here A
>T
is the heat trans-er area, 4
?,ottom
is the
len.th -rom the bottom o- the pipe to the lo(er
boundary o- a discretization, 4
?top
is the len.th -rom
the bottom o- the pipe to the top boundary o- a dis"
cretization, 7
li+
is the heat trans-er coe--icient (hen in
contact (ith li2uid and 7
gas
is the heat trans-er coe--i"
cient (hen in contact (ith the .as/ ;ote that the
(ei.hted heat trans-er area is a linearization o- the
inner pipe area -raction around the middle o- a cir"
cum-erential section and should thus only be used -or
relati3ely hi.h discretization 45F (ere considered
su--icient6/
Ahen calculatin. the pipe bendin. only the len.th
chan.e at the top 4element , and E in >i.ure E6 and
bottom 45 and + in >i.ure E6 o- the pipe is taken into
consideration/ The len.th chan.e is calculated
throu.h=
A4=4o(T
wall
)
4,6
(here o is the linear e1pansion coe--icient o- the
material/
4ift =rh 4E6
r=
4/ #
sin(o)
and sin(o)=
&
d
4F6
*igure .: Circumferential discreti&ation of the wall
#defined ,y user$
%
5
#
$
, E
F
+
*igure 9: @ipe ,ending
d
z
M
M
r
'
'i-t
h
H. Tummescheit, K. Tuszynski, P. Arnold
The Modelica Association 698 Modelica 2008, March 3
rd
4
th
, 2008
h=
4/ #
tan (o)
4%6
#.$, #3$ and #;$ yields,
4ift =
4
#
(
d
&

5
tan( o)
)
4+6
I- the li-t (ould reach hi.h 3alues o- se3eral cen"
timeters, the in-luence o- the li-t on the local le3el
and heat trans-er (ould ha3e to be taken into ac"
count, but such 3alues are outside o- the permissible
ran.e any(ays/
The (orst case encountered in the -inal modi-ied
3ersion o- the plant dia.ram (as -or a dead end o-
sli.hly less than , m len.th and a -illin. time -orm
empty to -ull o- about 55 minutes/ The (orst li-t (as
#/+ cm, a tolerable amount, and the (orst case stress
ratio usin. an e2ui3alent stress -rom the -ull three"di"
mensional stress tensor (as around &/,E/ The len.th
o- the pipe has the (orst e--ect on bo(in. as it e-"
-ects both the .eometry and the e1posure time, and
dead end pipes lon.er than , m (ould 2uickly cause
inacceptable bo(in./
2 Com%utational effort
>or cooldo(n scenarios o- the lar.er plant se.ments,
the computational e--ort (as 3ery hi.h= -or the most
comple1 se.ments o- the plant cooldo(n, *PL"times
o- $", days (ere necessary -or each simulation case,
and most o- the (ork is spent durin. the -irst -e(
seconds o- simulation time/ 7ymolaNs 3ersion o- the
dassl sol3er only mana.ed to sur3i3e the initial time
(ithout error (hen the option Ge2uidistant output
.ridH (as s(itched o--/ This in turn lead to result
-iles o- around 5)B that could not be handled by
7ymola and made postprocessin. 3ery tedious/
:3erall, -or system simulations o- the le3el o- com"
ple1ity encountered durin. the cool do(n simula"
tions, Ae see the -ollo(in. tool re2uirements -or
lar.e scale system simulations (ith short periods o-
3ery sharp .radients=
5/ A -ine .rained control o3er ho( many 3ari"
ables are stored and ho( o-ten they are
stored that does not in-luence the step"size
control al.orithm/ 7ymolaNs 7assl is a bad
e1ample o- a sol3er that takes the stora.e in"
ter3al into account in a (ay that lets simula"
tions -ail -or a small step size to stora.e in"
ter3al ratio/
#/ Means to in-luence step size control durin.
e1treme .radients under short time, or set a
minimum step size and .et (arnin.s in the
lo. (hen the re2uested accuracy (as not
achie3ed/
$/ Impro3ed numerical debu..in. -acilities/
7ymolaNs current debu..in. -acilities -or nu"
merical problems in lar.e models are insu--i"
cient/
* Conclusions
Modelica is not primarily kno(n -or its stren.th -or
modelin. partial di--erential e2uations, but due to its
suitability -or system le3el simulations, there are sit"
uations in (hich Modelica and 7ymola are an e1cel"
lent tool e3en -or models that re2uire a -ull $"dimen"
sional P7! discretization, under the contraint that
this only (orks -or simple .eometries/ In particular
the heat conduction e2uation (ith its simple struc"
ture can be combined (ith 5"dimensional t(o phase
-lo( -or thermal stress calculations/ The key ad3an"
ta.e is that it is possible to capture the most critical
thermal stress situation (ithin a comple1 plant (ith"
out the need to resort to co"simulation, or di--icult to
assess assumptions/
This simulation study re.ardin. cool do(n o- a cryo"
.enic trans-er system (as able to achie3e a number
o- .oals, in part because simulation (as used already
early in the desi.n process=
5/ It (as possible to establish desi.n .uidelines
re.ardin. dead pipe ends at closed branches
o- the net(ork to a3oid pipe bo(in./ The
.uidelines (ere incorporated in later re3i"
sions o- the desi.n/
#/ >lo( rates (ere optimized (ith respect to
the contradictory .oals o- minimum cryo.en
consumption and a3oidance o- strati-ied
-lo( conditions/
$/ imulation results allo(ed to de3ise cool
do(n se2uences that substantially decreased
the thermal stress -or all parts in the plant
e1cept the parts closest to the tanker used -or
-illin./
There are situations in (hich there is no possibility
to 3alidate simulations a.ainst measurements/ In
spite o- that shortcomin., simulation .i3es important
insi.ht into system beha3iour and e3en allo(s to im"
pro3e both system desi.n and system operation/
!3en 2uantitati3e analysis is possible to a certain de"
.ree (hen important parameters are (ell understood
and a care-ul sensiti3ity analysis is conducted (ith
respect to such parameters/
*ryo.enic plant simulations, e3en under the 3iolent
transients that occur durin. cool do(n o- trans-er
Simulation of Peak Stresses and Bowing Phenomena during the Cool Down of a Cryogenic Transfer System
The Modelica Association 699 Modelica 2008, March 3
rd
4
th
, 2008
lines, can be modeled easily (ith the cryo.enic op"
tion o- the Air*onditionin. 'ibrary/
3eferences
?5@ S%an. 34. Multiparameter !2uations o-
tate/ An Accurate ource o- Thermodynamic
Property 7ata, prin.er, Berlin, #&&&/
?#@ Tummescheit. /45 ?esign and <mplementa-
tion of A,Bect-Ariented Codel 4i,raries using
Codelica, 7issertation, T>9T"5&F$"!, 7epart"
ment o- Automatic *ontrol, 'und Inst/ o- Tech"
nolo.y, 'und, ch(eden, #&&#/
?$@ Tummescheit. /4. (born. 64 und Pr7l8. +45
Air*onditionin. < a Modelica 'ibrary -or 7yna"
mic imulation o- A* ystems, in Proceedings
of the 9th <nternational Codelica Conference,
Hambur., pp/ 5+E < 5I#, #&&E/
?,@ 9DI:;esellschaft 9erfahrenstechnik und
Chemiewesen 4!ditor6, D7I"AOrmeatlas, I
th
!di"
tion, prin.er, Berlin, #&&#
?E@ 9ersteeg. /4 +4 and alalasekera. </, An In"
troduction to *omputational >luid 7ynamics < The
>inite Dolume Method, Prentice Hall, 5IIE/
?F@ $emmon. (4, The 9e-Prop Lser manual, 3ersion
%/5, Personal communication ee also in-ormation
on 3ersion %/& at http=JJ(((/nist/.o3JsrdJnist#$/htm
4accessed #&&E"55"5E6/
?%@ "au%le 64/4, "isher "4(4, !n.ineerin. 7e"
si.n"A ynthesis o- tress Analysis and Material
!n.ineerin., Ailey, ;e( Cork, 5I+5/
?+@ http=JJ*ryo.enics/nist/.o3/ Accessed :ctober
#&&E/
?+@ =ounglo>e. B4A4, Thermophysical Properties
o- >luids/ I/ Ar.on, !thylene, Parahydro.en, ;i"
tro.en, ;itro.en Tri-luoride, and :1y.en, K/
Phys/ *hem/ 9e-/ 7ata, Dol/ 55, uppl/ 5, pp/
5"55, 5I+#/
?I@ D4/4 $iebenberg. 64+4 ?o>ak. "464 (des:
kuty= *ooldo(n o- *ryo.enic Trans-er ystems,
AIAA Paper ;o/ F%",%E/
H. Tummescheit, K. Tuszynski, P. Arnold
The Modelica Association 700 Modelica 2008, March 3
rd
4
th
, 2008