the other positions in the shield. The worst case is
As is evident, there is good agreement be-
the 19961997 winter at the bottom corner. Here
tween the numerical simulations and the actual
the difference between actual and numerical
temperatures recorded in the ledge material.
starts out at about 2C and converges to about
In addition to maximum frost penetration,
1.25C. The bottom corner agreement can be dra-
actual temperatures at several points around the
shield were compared with the simulation dur-
matically affected by the close proximity of the
ing the coldest time of the year. The thermocou-
ledge. In the numerical simulation, the ledge was
ples used in the verification as shown in Figure 4
modeled as smooth, uniform, and a constant dis-
are the top center inside (TC 19) and outside (TC
tance from the shield, at the bottom surface of the
21), the side inside (TC 22) and outside (TC 32),
trench. Obviously this is not the case in the actual
and the bottom inside corner (TC 23). Figure 14
construction, and this could account for some
shows the comparison between the actual tem-
temperature variations at the corner.
peratures and the simulation temperatures for
winter 19951996, and Figure 15 shows the com-
WENTWORTH STREET RESULTS
parisons for winter 19961997. The numerical
simulation generally produces very good agree-
The Wentworth Street site does not have a con-
ment with the recorded temperatures for both
trol string to measure the frost penetration, so for
years. The numerical temperature at the outside
the verification only temperatures in several loca-
middle of the top of the shield shows some dis-
tions around the shield were compared. Thermo-
crepancy at the coldest dips in the temperature
couples in the same relative location as at the La-
time record during both years. In the winter of
bossiere site were used and they are, as shown in
19951996, this amounted to about 2.5C at the
Figure 8, top center inside and outside TCs 6 and
5, side inside and outside TCs 8 and 12 and the
first cold dip around 7200 hr but decreased to
about 1.5C by 8000 hr. It appears that the numer-
inside bottom corner TC 11. Figure 16 shows the
agreement between the actual and simulated
ical simulation remained at the phase-change
temperatures. Again, the worst agreement comes
temperature longer than the actual temperatures
did between 6100 hr and 7100 hr and thus the
at the top center of the shield, which is also the
simulation did not reach the coldest temperatures
area that had the worst agreement using the La-
at 7200 hr. In the 19961997 simulation, the
bossiere data. Even here though, the general pat-
numerical dipped below the phase-change tem-
tern and timing of the simulation to the actual
perature before the actual temperatures, but the
data is very good, with the largest discrepancy
magnitude of the discrepancy at the coldest time
coming at the coldest temperatures around 7100
is only about 1C.
hr, where the difference is about 2.5C, and 7400
hr, where the difference is about 1.25C. Again the
This discrepancy in the top center outside ther-
mocouple (TC 21) was also noticed in the Went-
simulation appears to remain at the phase-
change temperature longer than the actual tem-
worth Street data. One possibility for this differ-
peratures did from about 6700 hr to 7000 hr. This
ence might be that the numerical simulation as-
could be related to moisture variation other than
sumes the moisture to be spread evenly through-
actual or to a time step refinement as explained
out the material when in the actual situation
above in the Labossiere Street data. The agree-
moisture migrates to the freezing front. The hori-
ment at the other locations is excellent, with any
zontal insulation boards block this natural mois-
errors generally well below 0.5C.
ture migration, so the area right above the shield
Since the actual physical configuration at
would not be receiving moisture from below. This
Wentworth was different from the design, i.e.,
area would therefore be drier than what the nu-
ledge did not continue to the surface, numerical
merical model assumes, which may result in low-
simulations were done to determine what the
er temperatures in this area than the numerical
performance of the actual shield would be under
prediction would give.
the design temperatures. In addition, the actual
It has also been noticed that a much smaller
water temperatures for Wentworth were not
time step in the FE program can have a substan-
known when the design was initially done. Ap-
tial impact upon the delineation of the phase
proximate temperatures were used that were ob-
front. A small time step, however, will substan-
tially increase the numerical simulation run time
tained by adding a factor to the measured 2nd
to obtain an accuracy that is not always necessary.
Avenue temperatures. The time-to-freeze proce-
The simulation shows excellent agreement at
dure was similar to that described earlier, except
14
Previous Page
