3. Benchmark
We employed three processes in the validation of the implemented
model. First, we compared results of time varying phase
front distributions from our full system model against Zhang and
Faghri [10] to verify that the semi-analytic approach was properly
executed. As such, the comparative results are unremarkable and
are omitted from this paper. Secondly, we compared against a separate
numerical study [6] for melting, whose own work was validated
with experimental data. While numerous numerical
studies exist, Lacroix [6] provides timewise variation of the heat
transfer rate, which has the most direct importance for application
rather than, e.g., PCM temperature profiles. We compared the total
transient heat transfer rate for three different constant HTF inlet
temperatures, which are defined relative to the melting temperature
of the PCM, n-octadecane. The purpose of the Lacroix [6] comparison
was to establish the accuracy of the semi-analytic
approach in isolation from the full system behavior. A time step
of 2 s was used with a discretization size corresponding to 317 elements—each
was selected by refining until solution independence
was achieved, which is the procedure used for all cases presented
within this paper. Lastly, we compared against full system data
from a representative hot battery experimental system we have
developed.
As shown in Fig. 4, the agreement with Lacroix [6] is very good
for the two cases corresponding to inlet temperatures within 10 K
of the melting temperature. However, for the 20 K case, we underestimate
heat transfer, relative to Lacroix [6], by roughly 10% during
the majority of the melting process. As a result, the melt front
reaches the wall at later time, so the transition to pure sensible
heat is delayed. The shape of the curve during the transition to sensible
and after is in reasonable agreement, albeit steeper for our
case. We suspect the reason for the discrepancy at higher inlet
temperature to be due to Lacroix’s [6] treatment of natural convection
in the PCM, which is based on an effective thermal conductivity
using the following empirical relation where coefficients were
determined to fit theoretical results to experimental data, as keff/
kPCM,l = 0.099 Ra0.25. Since the Rayleigh number is (Tw–Tm)
3
, it is
clear that increasing Tw–Tm will increase the effective thermal conductivity.
In addition, because a higher inlet temperature will result
in a higher mean wall temperature, the 20 K case will
similarly have a higher effective thermal conductivity. Therefore,
it seems reasonable to suggest that the axial convection, as modeled
by Lacroix [6], was not significant until the inlet temperature
was raised to 20 K above the melt temperature. While natural convection
is known to have an important contribution under particular
operating conditions [12], modeling for an actual multi-tube
arrangement would require consideration of tube-to-tube thermal
gradients, which simply is not practical computationally. Regardless,
the remainder of work presented in this paper considers solidification,
i.e., natural convection may safely be neglected.
For a full-system benchmark, we have developed an experimental
set-up for hot battery characterization using design input from
modeling. For the purposes of providing a benchmark, we performed
tests using a paraffin wax, whose latent heat and solidification
temperature we determined by differential scanning
calorimetry (DSC) analysis to be 187 J/g and 66 C, respectively.
3. BenchmarkWe employed three processes in the validation of the implementedmodel. First, we compared results of time varying phasefront distributions from our full system model against Zhang andFaghri [10] to verify that the semi-analytic approach was properlyexecuted. As such, the comparative results are unremarkable andare omitted from this paper. Secondly, we compared against a separatenumerical study [6] for melting, whose own work was validatedwith experimental data. While numerous numericalstudies exist, Lacroix [6] provides timewise variation of the heattransfer rate, which has the most direct importance for applicationrather than, e.g., PCM temperature profiles. We compared the totaltransient heat transfer rate for three different constant HTF inlettemperatures, which are defined relative to the melting temperatureof the PCM, n-octadecane. The purpose of the Lacroix [6] comparisonwas to establish the accuracy of the semi-analyticapproach in isolation from the full system behavior. A time stepof 2 s was used with a discretization size corresponding to 317 elements—eachwas selected by refining until solution independencewas achieved, which is the procedure used for all cases presentedwithin this paper. Lastly, we compared against full system datafrom a representative hot battery experimental system we havedeveloped.As shown in Fig. 4, the agreement with Lacroix [6] is very goodfor the two cases corresponding to inlet temperatures within 10 Kof the melting temperature. However, for the 20 K case, we underestimateheat transfer, relative to Lacroix [6], by roughly 10% duringthe majority of the melting process. As a result, the melt frontreaches the wall at later time, so the transition to pure sensibleheat is delayed. The shape of the curve during the transition to sensibleand after is in reasonable agreement, albeit steeper for ourcase. We suspect the reason for the discrepancy at higher inlettemperature to be due to Lacroix’s [6] treatment of natural convectionin the PCM, which is based on an effective thermal conductivityusing the following empirical relation where coefficients weredetermined to fit theoretical results to experimental data, as keff/kPCM,l = 0.099 Ra0.25. Since the Rayleigh number is (Tw–Tm)3, it isclear that increasing Tw–Tm will increase the effective thermal conductivity.In addition, because a higher inlet temperature will resultin a higher mean wall temperature, the 20 K case willsimilarly have a higher effective thermal conductivity. Therefore,it seems reasonable to suggest that the axial convection, as modeledby Lacroix [6], was not significant until the inlet temperaturewas raised to 20 K above the melt temperature. While natural convectionis known to have an important contribution under particularoperating conditions [12], modeling for an actual multi-tubearrangement would require consideration of tube-to-tube thermalgradients, which simply is not practical computationally. Regardless,the remainder of work presented in this paper considers solidification,i.e., natural convection may safely be neglected.For a full-system benchmark, we have developed an experimentalset-up for hot battery characterization using design input frommodeling. For the purposes of providing a benchmark, we performedtests using a paraffin wax, whose latent heat and solidificationtemperature we determined by differential scanningcalorimetry (DSC) analysis to be 187 J/g and 66 C, respectively.
การแปล กรุณารอสักครู่..
