Get documents about "Filter pressure drop as a function of flow velocity
Document Sample
........................................................................................................
Filter pressure drop as a
function of flow velocity for
SIVEX FC Aluminium Filters
Abstract The pressure drop/thickness data was converted to a
The object of this development was to obtain filter standardized format, and Darcian and non-Darcian
pressure drop as a function of velocity for the entire permeability coefficients were calculated. These
range of SIVEX FC foam filters. A secondary coefficients were input to MAGMASOFT, and several
objective was to validate this information using computer simulations were conducted to compare
computer simulation. The final objective was to the results with previous results obtained using the
incorporate this information into a database such “generic” filter information in the program. The
that foundries that use computer simulation could results were encouraging. Using the new filter data,
accurately model Foseco filters. All of these the user is able to more accurately predict the time
objectives were successfully met. to fill for the casting cavity. This is the ultimate
reason for having accurate filter pressure drop data
Several hundred filters were tested for pressure in computer simulations and the data has been input
drop characteristics using the water flow apparatus to MAGMASOFT.
at CINVESTAV, Saltillo, Mexico.
Introduction
All filters showed some variation in pressure drop The objective of this project was to obtain filter
that could be directly correlated with the foam pressure drop data as a function of velocity for the
structure. Regardless, in all cases, accurate entire range of SIVEX FC aluminum filters. The data
correlations were developed. These correlations was obtained via water modelling, and can be used
depended upon the porosity and thickness of the in current computer simulation software, as well as
filter only. In general, the pressure drop/thickness for internal performance evaluations.
increased with increasing porosity.
This report documents the pressure drop results for
all 22mm thick, 10, 20 and 30ppi SIVEX FC filters
(figure 1).
Figure 1: Filter data correlations
15
........................................................................................................
Method of Analysis Two different casting configurations were
The experimental setup, shown in Figure 2, evaluated:
consists of a continuous conduit in which water is
forced to flow through a test chamber. The water ❑ a test casting produced for an AFS gating and
flow rate can be varied, and is regulated by a risering competition
centrifugal pump.
❑ a truck wheel hub.
Figure 3 is the computer simulation model of the
test casting produced for the AFS competition using
a 2 x 6" KALPUR unit.
Figure 2: Schematic of experimental setup
The water flow is varied using three control valves,
and is measured with an electronic flowmeter. The
flow velocity test range was from 0.05 to 0.6 m/s,
which represents the likely flow velocity range for
most gating systems.
Once flow velocity reaches equilibrium for a given
filter, a differential manometer is used to measure
the pressure drop. The differential between the
manometer tubes is measured three times with a
digital Vernier, and an average value of ∆h is used
to determine the pressure drop. Pressure drop is
simply defined as:
Dp = ∆p g ∆h Figure 3: AFS test casting with 2 x 6” KALPUR
Results and discussion The casting weighs approximately 3.5 kg. The
The actual pressure drop/thickness data, and the pouring rate was approximated at 3.3 kg/s, which
corresponding correlations for SIVEX FC foam translates into an approximate fill time of 5 seconds.
filters can be seen and this information is used to To simulate the filling of this casting, we ramped
directly compare filter performance, and can be the inlet metal pressure from 0 mbar at 0 seconds
input to computer simulations to predict the flow to 8.92 mbar at 0.3 seconds, then maintained a
through a filter. constant 8.92 mbar for the rest of the filling cycle.
Pouring stream area and velocity of the metal (head
A second section is included to discuss how more height) were used to iterate to these pressure
accurate filter data invariably leads to more accurate conditions. These conditions seem to match actual
computer simulations of mould filling. foundry practice very well, as evidenced by the
results given below.
While x-ray comparisons were not made explicitly
for these castings, it should be noted that the A comparison was made between the standard
author has conducted significant work comparing MAGMASOFT 10ppi filter data which is a generic
simulation results using accurate pressure drop data preliminary dataset generated by industry, and the
to x-ray results (see references). In all cases, newly developed SIVEX FC 10ppi filter data. Both
simulations conducted using accurate pressure drop versions filled in approximately 5.2 seconds, which
data provide accurate flow characteristic results, is very close to the actual fill time of 5 seconds.
when compared to x-ray.
While the fill time predictions were very similar
Computer Simulations between the filter datasets, the SIVEX FC flow
Several casting simulations were conducted to predictions were more realistic than the generic filter
validate the SIVEX FC foam filter pressure drop data. results. Figure 4 shows both casting configurations
Comparative simulations were conducted to evaluate at 10% filled. Note that the SIVEX FC filter is already
the predicted fill time when using the standard, primed and flowing at this stage, while the generic
MAGMASOFT filter dataset for foam filters and when filter has just started to pass metal through the filter
using the newly developed SIVEX FC filter data. (SIVEX FC results shown in left hand picture.)
16
........................................................................................................
Figure 4: SIVEX FC vs Generic 10ppi Foam Filter at 10% Filled
The colors indicate velocity (greenish blue is 0 m/s, yellow/white is 1.2 m/s).
At 20 % filled (figure 5), the difference is even more dramatic. For the SIVEX case, more metal is passing through the filter, the KALPUR unit is
not backing up as fast, and the casting is filling differently as compared to the Generic results.
Figure 5:SIVEX FC vs Generic 10 ppi Foam Filter at 20% Filled
At 55% (figure 6), the story is the same. Striking differences can be seen between the velocities in the KALPUR unit and through the filter.
Figure 6:SIVEX FC vs Generic 10 ppi Foam Filter at 55% Filled
17
........................................................................................................
Figure 7: SIVEX FC vs Generic 10 ppi Foam Filter at 90% Filled
Finally, at 90% (figure 7), filled, the SIVEX FC case shows the more realistic filling cycle in which the casting is completely filled, and only the
top third of the KALPUR unit is yet to be filled with metal. The generic filter shows an unrealistic situation where the KALPUR unit is completely
full, while the casting will receive the last 10% of metal fill.
So, for this case, the generic filter is too restrictive and causes the KALPUR unit to overfill. It is important to point out these differences because
one of the most powerful applications of this filter data is to predict and analyze the filter’s effect on flow. Clearly, the SIVEX FC filter data is
more accurately representing the actual filling profile.
To get a better understanding of the filter data’s effect on flow, a close up of the filter was analyzed and showed significant differences
between the filter data sets. Figure 8 shows the flow velocities and vectors of the SIVEX FC and generic foam filters at 0.6 seconds into the fill.
The filters are cross-sectioned directly through the middle so that one can view the flow characteristics in the center of the filter. Again, colors
represent flow velocity (greenish blue is 0 m/s, yellow/white is 1.5 m/s).
Figure 8: SIVEX FC vs Generic 10 ppi Foam Filter at 0.6 seconds
The metal flow characteristics through the filter were significantly different between the SIVEX FC and the generic filter. At 0.6 seconds into
the filling, both filters have primed, but the SIVEX is allowing more metal through the filter, and the metal is not backing up as much as in the
generic case. This is because the SIVEX filter data is more realistically modelling the tangential, as well as the streamwise flow characteristics of
the filter, while the generic filter data is too restrictive. This is the case throughout the fill cycle for this configuration.
Other unrealistic effects shown in Figure 8 include eddy currents above the filter, entrapped air below the filter, and uniform velocity from end
to end within the filter. The SIVEX data is more realistic in these areas, and provides more accurate predictions of metal flow in both the
streamwise and tangential directions.
The simulations clearly show that the SIVEX FC and generic filter cases predict the filling characteristics of the mould cavity very differently.
However, for this case, the predicted porosity results are very similar. This is shown in Figure 9.
18
........................................................................................................
Figure 9: Porosity Predictions
For the most part, the simulation predicts porosity in the same locations for both configurations. So, for this example, the different filling
predictions did not significantly affect the order of solidification of the casting. Note that this is not always the case.
Wheel Hub Casting
Figure 10 is the computer simulation model of the truck wheel hub casting made using a 2x2" KALPUR unit. The casting weighs approximately
27 kg. The pouring rate was approximated at 3.7 kg/s, which translates into an approximate fill time of about 7 seconds. To simulate the filling
of this casting, we ramped the inlet metal pressure from 0 mbar at 0 seconds to 120 mbar at 0.3 seconds, then maintained a constant 120
mbar for the rest of the filling cycle. Pouring stream area and velocity of the metal (head height) were used to iterate to these pressure
conditions. These conditions seem to match actual foundry practice very well, as evidenced by the results given below.
Figure 10: Truck Wheel Hub Configuration with 2x2" KALPUR
The filling simulation results are also remarkably different. In all cases,
the SIVEX FC flow predictions were more realistic than the generic
filter results. Figure 11 shows both casting configurations at 10%
filled. Note that the SIVEX FC filter is already primed and flowing at
this stage, while the generic filter has just started to pass metal
through the filter.
The colors indicate velocity (greenish blue is 0 m/s, yellow/white is 2.0
m/s).
At 10% filled (figure 11), the SIVEX FC filter is allowing metal to flow
more freely through the filter, and the metal is beginning to enter the
casting cavity. The generic filter is more restrictive.
19
........................................................................................................
Figure 11: SIVEX FC vs Generic 10 ppi Foam Filter at 10% Filled
At 25% filled (figure 12), the generic filter case shows the pouring basin to be completely filled, while the SIVEX case is still only partially filled.
At this point, MAGMASOFT imposes a uniform boundary condition that prevents the metal from overspilling the pouring basin, and thus the
pouring pressure input is overwritten. This means that the generic filter case will artificially change the metal mass flow rate.
Figure 12: SIVEX FC vs Generic 10 ppi Foam Filter at 25% Filled
By 65% filled, the boundary conditions imposed on the generic filter case allow the filling profile predictions to catch up to the SIVEX case. As
shown in Figure 13, the filling profiles begin to look somewhat similar at 65% filled.
Figure 13: SIVEX FC vs Generic 10 ppi Foam Filter at 65% Filled
So, for this case as well, the generic filter is too restrictive and causes the pouring basin to overfill.
20
........................................................................................................
Figure 14: SIVEX FC vs Generic 10 ppi Foam Filter at 0.5 Seconds
A quick look at the details within the filters for this case is shown in Figure 14 at approximately 0.5 seconds into the fill. Again, colors
represent flow velocity (greenish blue is 0 m/s, yellow/white is 1.5 m/s).
Again, the metal flow characteristics are considerably different between the two filter datasets, with the generic dataset exhibiting unrealistic
flow characteristics above, within and below the filter. As stated above, this is believed to be because the SIVEX filter data more realistically
models the tangential, as well as the streamwise flow characteristics of the filter, while the generic filter data is too restrictive
Figure 15 shows an x-ray view of the porosity predictions for both cases. As before, the differences in filling profiles did not seem to alter the
predicted porosity, either in size or in location. This is odd considering the very different fill profiles, and the significant difference in fill time. As
stated earlier, this is not always the case. In our steel and iron filtration work, some differences in porosity prediction were realized when
comparing results generated using the FOSECO and generic filter datasets.
The predicted pour time results are closer to the actual pour time when using the new FOSECO filter data. It is likely that the longer the pour
time, the larger the difference between the two filter datasets, and thus the greater the error when using the standard filter data. The accuracy
of mould filling simulations should increase across the board with this new data, and this will be especially apparent for the larger pour times.
Figure 15: Porosity Predictions
Casting Actual Pour Time (sec) Predicted Pour Time w/ Predicted Pour Time w/
Foseco Filter (sec) Generic Filter (sec)
AFS Test Casting 5 5.3 5.2
Truck Wheel Hub 7 7.5 12.7
Table 1: Fill Time Comparisons
Table 1 summarizes the filling results. A comparison was made between the standard MAGMASOFT 10ppi filter data and the newly developed
SIVEX FC 10ppi filter data. For this case, there was a significant difference in fill time predictions. The SIVEX FC filter data yielded a fill time of
about 7.5 seconds, while the generic filter data resulted in a predicted fill time of 12.7 seconds.
21
........................................................................................................
Summary / Conclusions Acknowledgements
Pressure drop data was collected for SIVEX FC foam The author would like to thank FOSECO colleague
filters using the water flow facility at CINVESTAV, Brian Began who made significant contributions,
Saltillo, Mexico. both in categorizing the filters to be tested, and in
providing valuable details regarding filter production
Pressure drop/thickness correlations were developed and the air test pressure drop device. Also, Mairtin
for the entire product range of SIVEX FC foam Burns and Phil Dahlstrom should be recognized for
filters. Pressure drop for these filters was dependent conducting all of the computer simulations for the
upon porosity and filter thickness. validation section of this report. Finally, Dr. A.
Humberto Castillejos E., and Dr. F. Andres Acosta
Because the data was consistent, individual pressure G. Of CINVESTAV, Saltillo, Mexico are to be
drop/thickness correlations were developed for each commended for their outstanding work.
type of SIVEX FC filter, including 10, 20 and 30 ppi CINVESTAV tested each of the filters for this
filters. In general, the pressure drop/thickness values project, and provided quality, repeatable results that
increase with increasing porosity. allowed data correlations to be possible. In addition,
CINVESTAV provided valuable technical input to
The variation in foam, and thus filter geometry help us understand filter flow and pressure drop
(porosity), is minor. However, some filter to filter characteristics.
variation was confirmed using the water tests.
Further references
The pressure drop/thickness correlations were A.C. Midea, J. Outten, “FILTERCALC for Steel – A
converted to a standardized description of filter Windows Based Programme for Sizing Foam Filters
flow characteristics, which involves computation of for Steel”, Foundry Practice, Issue 240, June 2003.
the Darcian and non-Darcian permeability
coefficients. The power of these coefficients is that A.C. Midea, B.A. Alquist, G. Strauch, E. Wiese,
they are the standard for characterizing filter flow “Innovative Use of Computer Simulation and Real
performance, and are not dependent upon the fluid Time X-Ray Technology to Optimize Steel Gating
medium used to collect the pressure drop data. Systems”, Foundry Practice, Issue 239, June 2003.
More importantly, the pressure drop performance
of the filter can be determined for any fluid flow A.C. Midea, B.A. Alquist, G. Strauch, E. Wiese,
using these coefficients, assuming that the fluid “Innovative Product Development Using Virtual
viscosity and density at the pouring temperature are Planning and Simulation Delivers Greater Flexibility
known (see references). These are the values used in the Design of Gating Systems for Steel Castings”,
for computer simulation analyses. Presented at the GIFA Exhibition, Dusseldorf,
Germany, June 2003.
Computer simulations were conducted for three
casting configurations. Each configuration was run A.C. Midea, “Modelling of Cellular and Foam
with the appropriate FOSECO filter, and re-run with Filtration Devices in Iron Casting Simulation”,
the standard filter dataset in MAGMASOFT. The Presented at 65th Annual AFS Wisconsin Regional,
program was given the metal flow stream February 14-15th, 2002, Milwaukee, WI
information, and asked to predict the fill time. In all
cases, the actual cellular filter data provided a more A.C. Midea, B.A. Alquist, “Increasing the Accuracy
accurate answer than the standard dataset. of Metal Flow Results”, Foundry Management and
Technology, August 2001, SFSA Paper, Presented
This validates the accuracy and usefulness of the at National T&O Conference, November 1-3rd,
data for computer simulations. 2001, Chicago, IL
It should be noted that porosity predictions can be A.C. Midea, “Pressure Drop Characteristics of Iron
affected by the filter pressure drop data. More Filters”, AFS 01-042, Presented at 105th AFS
accurate filling predictions can result in more Casting Congress, April 2001.
accurate porosity predictions.
22
........................................................................................................
A. H. Castillejos E., F.A. Acosta G., “Fluid-Dynamic
Characterization of Ceramic Filters”, CINVESTAV,
Unidad, Saltillo, Mexico, (Foseco Contracted
Report), April 2000.
M.D.M. Innocentini, P. Sepulveda, V.R. Salvini, V.C.
Pandolfelli, “Permeability and Structure of Cellular
Ceramics: A Comparison Between Two Preparation
Techniques”, American Ceramic Society Journal,
Vol 81,No. 12, pp. 3349-3352, 1998.
M.D.M. Innocentini, V.R. Salvini, V.C. Pandolfelli,
J.R. Coury, “Assessment of Forchheimer’s Equation
to Predict the Permeability of Ceramic Foams”,
American Ceramic Society Journal, Vol 82, No. 7,
pp. 1945-1948, 1999.
M.D.M. Innocentini, V.R. Salvini, V.C. Pandolfelli,
J.R. Coury, “The Permeability of Ceramic Foams”,
The American Ceramic Society Bulletin, September
1999.
M.D.M. Innocentini, A.R.F. Pardo, V.R. Salvini, V.C.
Pandolfelli, “How Accurate is Darcy’s Law for
Refractories”, The American Ceramic Society
Bulletin, November 1999.
23
Related docs
