MICA Project
[SAMSON]
Author(s):
Stewart Miles and Suresh KumarOrganisation: Building Research Establishment – Fire Research Station (BRE-FRS)
Report date: May 1998
1 Introduction to the industrial problem(s)
The role of BRE-FRS as an end-user in the MICA project was to specify
the VR Interface and validate MICA for smoke movement and fire spread applications within
the built environment. In particular, the interests of building service designers, fire
safety consultants and the fire protection industry were to be considered. With the advent of increased desktop computing power, coupled with the
increased complexity of modern buildings, fire-modelling methods are in more widespread
use. The subject of ‘fire safety engineering’ has emerged in recent years both
for this reason and also due to a worldwide trend towards functional based building (fire)
regulations, which can be more appropriate for complex, large buildings than traditional
prescriptive codes. CFD fire models are probably the most sophisticated element of fire
safety engineering. They allow detailed predictions for the smoke and heat transport
processes associated with both ‘real life’ and ‘design’ fires. A
design fire may be typically a 5x106W fire source for which a proposed smoke
control system must operate successfully. By virtue of its complexity, academics and specialist engineers have
until recently been the main users of CFD. However, as the software has matured and the
required computing power become more widely available, interest in CFD has spread to
industry at large, include building design engineers and architects, for whom fire
modelling is an important requirement. However, despite the potential benefits of CFD, the building and fire
control industries are not exploiting the technology to its full potential. One main
reason is the steep ‘learning curve’ associated with traditional CFD programs.
Another is the impractical computer resource required to simulate ‘real life’
industrial problems (even with the advent of modern ‘high performance’ personal
computers). One solution to the problem is to appoint CFD consultants to undertake the modelling
task. Another, potentially more cost effective, solution would be to make the
CFD models simpler to use and to provide easy access to high performance computing
facilities. MICA, and the associated Internet network of computing centres MICANET, has
the potential to satisfy both these requirements. Drawing upon its own experience in fire safety engineering, BRE-FRS has
evaluated the current MICA software (PHOENICS VR and the Internet communication program
mica-loc) using the SAMSON Special-Purpose Program (SPP). Two fire-modelling scenarios were identified in Work Package 2. Two
further scenarios have been added since, providing four scenarios for the validation
exercise. These are described separately below. Although not exhaustive in their scope,
the four scenarios provide a cross-section of the type of problem that may be encountered
in ‘real life’. They cover a wide geometric range, and test at least some of the features that would be
required by a fire safety engineer. Important aspects that have been addressed include: All four scenarios were set up from scratch using the VR Editor. Steady
state and transient simulations have been performed at the remote computing centres.
Furthermore, ‘coarse grid’ simulations have been performed locally on the PC in
order to examine the effect of different relaxation parameters etc. Simulation results for
all scenarios have been viewed with the VR Viewer. All comments in the remainder of the BRE-FRS report relate specifically
to PHOENICS 3.1 and mica-loc, distributed to the industrial partners at the beginning of
February. Validation work had commenced with earlier releases of MICA, but where
appropriate the comments made at that time have been superseded with those for the latest
release. Emphasis is placed on the requirements of the MICA project, which is to validate
the concept of MICA from the point of view of potential industrial end- users.
2 Description of the simulated case(s)
The four scenarios studied by
BRE-FRS are described here. Two scenarios, the warehouse and environmental office, were
described in Work Package 2 (they have since been modified, but the general ideas remain
the same). The other two, the corridor fire experiment and the forced ventilated nuclear
cell, are new scenarios. They were added to provide a broader validation of MICA. 2.1 Warehouse Fire With External Dispersion This is a hypothetical example,
based loosely on a scenario studied by BRE-FRS in a previous European Commission project1
under the Environment Programme. The basic idea was to model the development of a fire
within a warehouse building and predict the dispersion of the fire products into the
surrounding environment. Two doorways and a set of roof vents provide ventilation, and a
wind profile is prescribed at one end of the domain. This type of example is of great interest to parties involved in the
building and siting of large buildings to store hazardous substances such as pesticides.
It is important to know the potential danger of fire at these sites, in particular with
regard to the houses and other buildings in the near vicinity. The interaction of the
prevailing wind pattern with the fire products leaving the building will determine safe
distances for adjacent properties. Figure 1 shows the geometry of the warehouse and the external
atmospheric domain. Each roof vent has dimensions 2.5m x 2.0m. Two fire sources are
included, both having a steady state heat release rate of 5.0x106W. In the
earlier EC study the fire was modelled as a pool fire of xylene, a common solvent for
agrochemical storage. However, the SAMSON SPP requires fires to be modelled currently as
volume heat sources. Two fires are included to make the case more interesting and
challenging. Figure 1 Plan and side views of the warehouse The wind inflow boundary is 50m upstream of the warehouse. In order
to capture the downstream flow patterns it is necessary to extend the domain a long way
downstream (10 times building height at the very least). Figure 2 External domain surrounding the warehouse Wind speed varies as a function of distance above ground according
to the following equation, where u = wind speed [ms-1], z = distance above ground [m], z0 = ground roughness parameter (height) [m], h = reference (free-stream) distance above ground [m], and subscript z and h denote distances z and h
above ground respectively. A more complete model would be transient with growing fire source(s)
and opening/closing vents and doors. However, for the purpose of MICA validation the vents
and doorways have been fixed open, and the fire (heat) sources have been defined so far as
steady volume sources (further work with transient sources may be performed before the
final project review). 2.2 Corridor Fire Experiment A corridor fire experiment2,
performed by the Swedish National Testing and Research Institute, was selected as the
third test case. The experiment consisted of a corridor (6m by 7.5m cross-section and 15m
long) inside a much larger burn hall. The scenario that was considered in the MICA project
has a growing propane fire in the centre of the corridor floor and one open roof vent 3m
from the fire. Figure 3 illustrates the geometry of the corridor and the location of
air inlets at the burn hall floor level (either side of the corridor apparatus). Note that
the ends of the corridor (i.e. west and east limits) are completely open. Figure 3 Geometry for the corridor fire experiment In the published paper the fire heat release rate for the test
modelled is given as, Click here [MW] where t is time from fire ignition. The published data from the experiment allowed the MICA results to be
validated. In particular, there are data for the heat release rate through the vent as a
function of time and also the temperature and velocity profiles of the ceiling jet. The actual scenario modelled for MICA was simplified slightly. A
symmetry plane was prescribed through the centre plane of the corridor, fire and ceiling
vent. Furthermore, the floor level air inlets were specified differently. Rather than
prescribing a fixed flow rate at the inlet, a fixed (zero) pressure boundary was set up 2m
below floor level so that volume of air entering through the inlet is driven by the
entrainment process at the heat source. Figure 4 shows the geometry of the scenario modelled in MICA. Figure 4 Plan and side views of the corridor fire experiment A volume heat source with dimensions 1m x 0.5m (floor area) x 2m
(high) and a 3.25x105W heat release rate represents the fire here. This corresponds to a
time four minutes into the test. Note that the ‘actual’ heat source, accounting
for symmetry, has twice this floor area and heat output. The corridor sides and ceiling
are assumed to be smooth and adiabatic The MICA work involved mainly steady state simulations. However some
transient simulations were performed in order to evaluate this aspect of MICA. 2.3 Forced Ventilated Nuclear Cell This example is based on a
published fire test in a forced ventilated nuclear cell enclosure at the Lawrence
Livermore National Laboratory. BRE-FRS studied it in the 1980s3 as a validation
case for its emerging CFD modelling capability. The enclosure has dimensions 4m x 6m x 4.5m, an inlet opening (2m x
0.12m), an extract fan (0.65m x 0.65m) and a pool fire (4x105W heat release
rate) in the centre of the floor. Figure 5 shows the geometry. For most of the MICA
simulations the domain was extended 2 metres beyond the inlet, with a zero pressure
boundary on the north face of the domain. Figure 5 Geometry for forced ventilated nuclear cell The fire was represented in MICA as a 0.8m x 0.8m (floor area) x
1.5m (high) heat source with a 400 kW heat release rate. The extract fan in the experiment had a volume flow rate of 0.4 m3s-1,
equivalent to a uniform outflow velocity of 0.947 ms-1. The problem was selected because it requires a ‘robust’
mechanism for transferring heat from the gas phase to the enclosure boundaries. In the
experiment, in excess of 75% of the heat released from the fire was absorbed by the
enclosure structure. Boundary heat loss is certainly very important in
‘flashover’ fires, and can also be important for general smoke movement
problems. Some new JASMINE simulations were undertaken during the course of the
validation exercise, both using a true fire source and the heat source approximation. Only steady state simulations have been considered, both with MICA and
JASMINE. 2.4 Environmental Office The scenario considered here
differs from the one proposed originally; a reduced geometry is considered and, although a
fire source is not included, multiple heat sources are present. These heat sources
represent people and personal computers (100W and 200W respectively). There are a number
of important issued related to heat transfer and buoyancy that can be addressed by these
heat sources. Furthermore, since in SAMSON fire sources are simply heat sou rces, it was
sufficient for initial validation, at least, to concentrate on the pre-fire scenario. The concept behind this scenario is taken form the new
‘Environment Building’ at BRE. Instead of using conventional air conditioning,
the building is designed to provide air control and cooling by natural ventilation. The
concrete ceiling and floor provide summer cooling of the air, and are at a lower
temperature than the outside (ambient) air. Ventilation is provided either by the
prevailing wind flow or by the use of external stacks. Airflow up the stacks is driven
either by a natural p ressure gradient or if required, is assisted by fans at the top of
the stacks. A hypothetical case has been examined in the MICA project in which the
ambient air is at 28ºC and the ceiling and floor are at 20ºC and 24ºC respectively. A
‘slice’ of the office has been modelled, making use of symmetry planes to mimic
the cyclic nature of the office. A zero speed wind condition has been specified, with air
being ‘pulled’ up the external stacks by fans, at a rate equivalent to two air
changes per hour inside the office. It should be noted that the mode of operation repre
sented by this scenario is a purely hypothetical one for the purposes of the MICA project. Figure 6 shows the geometry of the modelled ‘slice’, with the
internal objects (desks, heat sources) hidden from view. Air is drawn in through the three
open ceiling ducts, and exits through the two lower level windows into the stack. The area
to the left (low ‘x’ co-ordinate) is a computer suite, and is partially screened
from the rest of the office. All solids apart form the ceiling and floor are adiabatic.
Note that only one storey has been considered here. The first floor is essential ly a
repeat of the ground floor, but is ‘blocked-in’ for this work. The orange colour shows the ambient air that is cooled down as it
passes through the ducts. The air inside the office is then warmed by the heat sources,
and is extracted through the vertical stack. In the ‘slice’ model, there are two
openings from the office space into the stack. Note that because of symmetry, the ducts
and stack are actually twice as wide as shown in the plan view. The layout of desks, people and computers inside the simulated area is
shown in the VR Editor pictures in the pre-processing section. Figure 6 Plan and side views of the office ‘slice’ Although there is no experimental validation data for this
particular scenario, it has been modelled with the BRE-FRS CFD model JASMINE1,3,
with which the authors have extensive experience and confidence. The JASMINE data can
therefore be use to provide qualitative, if not quantitative, validation of the numerical
results from MICA. 3 Pre-processing of the case
3.1 Problem specification
This section describes how the four BRE-FRS test cases were set up and comments on the current state of PHOENICS VR with respect to setting up general fire and smoke movement problems. This covers the VR Editor, the availability of objects (fire sources etc) and the selection of physical models
The four test cases were set up from scratch using the SAMSON SPP. Figures 7-12 show VR Editor frsges for the cases. Various objects have been hidden to allow the scenarios to be better appreciated, and some annotation added. Note that some colours were corrupted in taking the bitmap frsges from the VR Editor window to the document. The colours used within the VR environment are more aesthetically pleasing!
The warehouse and corridor fire experiment was the most straightforward to set up. This was due, in part, to all solids being adiabatic. With the exception of the ground in the warehouse case, these solids were BLOCKAGE objects attributed to material 198 – "solid with smooth-wall friction". The ground was a PLATE object with a roughness of 10mm.
Non-adiabatic solids were included in the environmental office and forced ventilated nuclear cell scenarios. However, as reported in the CFD simulation section, these caused some difficulty. For the environmental office, the air inflow ducts and floor were at 8ºC and 4ºC below ambient respectively. Both ‘fully participating’ solids (i.e. conjugate heat transfer) and isothermal solid boundaries were investigated in the forced ventilated case, with isotherma l boundaries being preferred for the final simulations reported in this study.
Currently, fire sources must be specified as volume heat sources. These are BLOCKAGE objects attributed as ‘Air using Ideal Gas Law’ with a heat source, as in the following example, taken from the warehouse case
The k-e turbulence model has been used in nearly all the simulations performed.
Figure 7 Warehouse domain – wind boundary condition on left face
Figure 8 Warehouse close-up – openings, fires and null objects
The main challenge in setting up the warehouse case was the selection of the open boundaries. In earlier releases of PHOENICS VR a sequence of adjacent horizontal INLET objects were required in order to specify the required wind profile. However, PHOENICS 3.1 includes a new, more generalised INLET object in the SAMSON SPP, which enables a single INLET object to specify the entire wind boundary condition according to the log-law presented in Section 2.1. Th e attributes associated with the INLET object are as follows
All other domain boundaries, apart from the ground, were set up as zero-pressure boundaries. This was achieved using OUTLET objects with the following attributes
The warehouse case made good use of the ‘holes in walls’ feature, whereby once a roof or wall BLOCKAGE object has been created, a vent or door is superimposed by adding a ‘clear air’ BLOCKAGE consisting of ‘Air using Ideal Gas Law’.
The two heat sources were set to steady values of 5x106W for the steady state simulations performed so far (some transient simulations with a growing fire source may be performed before the Final Review Meeting).
NULL objects were added in order to refine the grid in the vertical direction inside the building and in the x-y plane at the heat sources and around the outside perimeter the building. NULL objects are discussed further in later sections.
Figure 9 Environmental office – internal objects and stack
Clip-art (the shape attribute) was used to more effect than in the environmental office scenario than in the other scenarios. In particular, the heat sources (computers and people) were shown realistically.
Again, zero-pressure boundaries were established using OUTLET objects. The ‘clear air’ BLOCKAGE feature was again used to create holes in walls.
Figure 10 Environmental office – air ducts revealed
At the top of the external stack there is an extract fan. This is implemented in the final version of the VR specification as a FAN object. An INLET object can be used instead, since FANs and INLETS are fundamentally the same (a FAN does not allow for log-law or power-law velocity profiles). In either case, the source term may be defined in terms of an inflow/outflow velocity or as a volume flow rate. The attribute dialogue box for the latter definition is shown below
Note that the term ‘Mass Flow Rate’ is confusing, since it refers in fact to a volume flow rate (in m3s-1). The temperature is that that is presumed to prevail at the FAN (or INLET) object, and for a non-isothermal fluid flow problem this is a limitation. This is discussed further in later sections.
The ceiling, ducts and floor were fixed temperature BLOCKAGE objects, as shown below in the attribute dialogue box for the ceiling slab
The air inlet ducts were created from slabs connecting at right angles below the ceiling. In earlier work ‘wedge’ clip-art shapes were used in order to approximate a sinusoidal ceiling more realistically. However, the ‘wedge’ clip-art is awkward to use, in particular in respect to orientating the shapes correctly. It was decided for the purpose of the current validation exercise that rectangular ducts were sufficient.
Figure 11 Corridor fire experiment
The corridor fire experiment was probably the simplest of the test cases. It was set up for both steady state and transient simulations. In both cases the heat release rate of the heat (fire) source was fixed at 3.25x105W.
Conversely, the forced ventilated fire scenario was, overall, the most difficult. In terms of geometry it was certainly straightforward. However, the nature of the problem, particularly the solid boundary heat transfer processes, made the CFD simulation very difficult within the confines of the SAMSON SPP.
The problem specification used in the final simulations employed isothermal solid boundaries, using BLOCKAGE objects as in the environmental office example. The extract fan was again specified as a FAN object, with a ‘volume’ flow rate of –0.4 m3s-1 and temperature of 250ºC (the approximate temperature measured in the original experiment).
A difficulty was encountered in using a combination of a ‘clear-air’ BLOCKAGE, a fixed temperature solid BLOCKAGE and a FAN object (discussed in the CFD simulation section). Hence, the wall was constructed from four separate segments surrounding the extract fan, which made the editing a little more involved.
Figure 12 Forced ventilated nuclear cell
Further, general observations concerning problem specification are provide below. Although most of these refer to additional features that are required/desired, it should be noted that the author found the basic problem specification concept to satisfy the requirements of MICA.
where hc = heat transfer coefficient [Wm-2 ºC-1]
Tg = temperature of the gas at the near-wall node [ºC]
Tw = temperature of the surface (wall) adjacent to the near-wall node [ºC].
The value of hc my typically be about 25 Wm-2 ºC-1, or may be a function of surface temperature.
3.2 Data conversion to PHOENICS input format
This aspect of MICA seems to be reasonably robust. The Q1 file generated by the VR Editor seems to be complete, and is easily edited by the more experienced user to either modify any settings prior to performing a simulation or to add a case-specific PIL fragment at the end of the file.
Further comments are now provided.
> OBJ5, NAME, FLOOR
> OBJ5, POSITION, 2.700000E+00, 0.000000E+00, 0.000000E+00
> OBJ5, SIZE, 1.375000E+01, 3.000000E+00, 1.000000E-01
> OBJ5, CLIPART, cube1
> OBJ5, ROTATION, 1
> OBJ5, TYPE, BLOCKAGE
> OBJ5, MATERIAL,
117> OBJ5, TEMPERATURE, 2.400000E+01
> OBJ5, INI_TEMP, 2.400000E+01
BUOYE=20.
relax(u1,falsdt,1.)
relax(v1,falsdt,1.)
relax(w1,falsdt,1.)
relax(ke,falsdt,10.)
relax(ep,falsdt,10.)
fiinit(u1)=2.
Obviously, the ultimate aim of MICA should be to eliminate the need for any case-specific PIL fragments since this is really for the more experienced PHOENICS user.
falsdt(u1) = 0.01
afalsdt(v1) "
falsdt(w1) "
falsdt(ke) = 0.001
afalsdt(ep) "
This was found, however, not to be adequate for good convergence. Hence the addition of PIL fragments as quoted in the previous comment.
Ultimately, the specification of relaxation parameters should be removed form MICA completely (except for an expert user), with either the local PHOENICS VR program or the remote processing centre handling all relaxation issues.
The automatic grid refinement algorithms being developed as part of the MICA project will go a long way towards satisfying this goal.
PATCH(OB4 ,WEST , 1, 1, 7, 8, 14, 15, 1, 1)
COVAL(OB4 ,P1 , FIXFLU ,
-6.307E-01)COVAL(OB4 ,U1 , 0.000E+00,
-9.467E-01)COVAL(OB4 ,V1 , 0.000E+00, 0.000E+00)
COVAL(OB4 ,W1 , 0.000E+00, 0.000E+00)
COVAL(OB4 ,KE , 0.000E+00, 9.946E-04)
COVAL(OB4 ,EP , 0.000E+00, 1.583E-04)
COVAL(OB4 ,TEM1, 0.000E+00,
2.500E+02)A
mass source is defined according to the volume flow rate (or equivalent velocity definition), the material (gas) property and the temperature attribute. This mass source is subsequently fixed regardless of the local fluid conditions in the simulation. Hence a fixed volume flow rate is not achieved for a non-isothermal flow.
4 Submission of the case, remote calculation and data retrieval
4.1
Job submission and Internet communicationThe latest version of the job submission and Internet communication program (mica-loc) seems to have removed the main problems encountered with the early releases. Most importantly, mica-loc is able now to communicate through proxy firewalls. This does however require the firewall system administrator to configure the necessary communication ports.
BRE-FRS submitted 31 remote simulations between the end of February and the middle of April, using PHOENICS 3.1 and the associated mica-loc Internet communication program. Most of these were sent to the CHAM, with the remainder sent to Inria. The breakdown of this figure between the four test cases is:
warehouse fire _ 9
corridor fire experiment _ 5
forced ventilated nuclear cell _ 11
environmental office _ 6
Not all these simulations achieved a solution, and the early ones were really just test cases to try various things out. Each simulation typically involved about 1000 iterations with a grid of about 40000 cells, and was completed in 6 to 24 hours.
Other comments are provided below:
4.2 Expert Intervention
This feature was not used explicitly by BRE-FRS. However, expert intervention was performed at CHAM for a number of simulations where it had been noticed that the simulation was having ‘problems’. This was certainly a useful service.
Even if a MICA client does not request expert intervention, some form of expert ‘supervision’ is of great benefit. This could be either ‘manually’ by experts at the remote centres or ‘automatically’ by some form of convergence checking routines. The latter, however, would require a lot of further work.
4.3 Data retrieval
Data retrieval is closely related job submission and Internet communication. Additional comments, specific to obtaining simulation results, are:
5 Post-processing of the case
Comments about the VR Editor, made in the pre-processing section, apply also here. Furthermore, the CFD simulation section includes graphic representations of the results viewing capability. Comments specific to the viewing and analysis of the CFD results is provided below.
Figure 13 Pollutant iso-surface for the warehouse fire
6 The CFD simulation
CFD simulations were performed both locally on PC and, more importantly, remotely at the MICANET centres. Much of the investigation into the correct setting of boundary conditions, initial values and relaxation parameters was achieved with relatively coarse grid (~2000_10000 cells) solutions on the PC. More detailed results using finer grids (~20000_80000 cells) were then obtained by remote simulation.
The CFD simulations have highlighted some important points, outlined below:
The simulation of the four test cases is now described, and various comments and recommendations are made. The findings are summarised more concisely in Sections 7 and 8.
6.1 Corridor fire experiment
Of the four test cases selected by BRE-FRS, the corridor fire experiment was the simplest to simulate. A realistic flow pattern could be obtained fairly easily. Although most effort was directed towards the steady state simulation, some transient simulations were performed with the heat release rate of the fire (heat) source fixed at the steady value.
Figure 14 shows a VR Viewer frsge of temperature vectors on the symmetry plane of the corridor fire case. The simulation was performed remotely using a grid of 36x28x24 cells and 1000 iterations. Unfortunately, once again some colours have been corrupted in the transfer form the VR Viewer window to the Word document, and hence the temperature colour map is a bit confusing.
Figure 14 Temperature vectors for the corridor fire experiment
The flow pattern appears correct, with flow out of the vent (hidden in the figure) and around the two ends of the corridor. However, the temperature field away from the fire plume is too low. At the probe location we see a temperature rise of about 27ºC, whereas in the published experimental data the temperature rise at this location is about 42ºC. It seems that too much of the heat is localised in the vertical plume. This will be due, in part at least, to using a volume heat source instead of a ‘true’ fire source. In the actual experiment the plume would have been more ‘diffused’, and may have leaned sideways slightly.
For the record, the residuals after the 1000 iterations were as follows:
Whole-field residuals before solution
with resref values determined by EARTH
& resfac= 1.000E-02
variable resref (res sum)/resref
P1 4.278E-03 4.975E+00
U1 4.014E-04 9.933E+01
V1 3.288E-04 8.058E+01
W1 1.085E-03 4.930E+01
KE 5.591E-05 1.944E+03
EP 1.230E-05 1.152E+04
TEM1 1.967E+02 6.925E-01
A remote transient simulation was performed also for the corridor fire experiment. This involved the following run parameters:
The remote simulation generated three ‘phi’ files (at 10s, 20s and 30s). The evolution of the scenario could be observed easily. Figures 14 and 15 show temperature vectors at 10s and 30s respectively. Note that the probe is here located at the vent. – removed to make the report more compact
The content of the result data generated by MICA for a transient simulation was commented on in the remote data retrieval section.
6.2 Warehouse fire with external dispersion
The warehouse fire scenario was of particular interest because of its combination of internal and external flow regimes. Initially, this proved to be a difficult problem to converge. However, once the necessary case-specific PIL fragment had been added to the Q1 file, some sensible results were achieved.
The most important case-specific PIL instruction was the setting of the initial u1 velocity component (wind direction) to the free-stream value, which was 2 ms-1 for the results reported here. Without initialising the u1 field, the flow pattern immediately diverged, with air leaving the domain through the upper open (pressure) boundary. Once this happened, it was impossible to ‘correct’ the flow.
A second aspect of the case-specific PIL fragment used in the warehouse case is the re-setting of the relaxation parameters for the momentum and turbulence equations. As reported above, a single relaxation parameter, specified in the SAMSON ‘solution options’ dialogue box, is too general. For the warehouse case it was found beneficial to employ smaller false time-steps for the momentum equations compared to the turbulence equations (which is the reverse of the default SAMSON setting).
Figure 15 shows the monitor and residual plots for a local PC simulation of the warehouse case using a 25x25x15 grid and false time-steps of 1.0 for the momentum equations and 10.0 for the turbulence equations. The monitor location is above one of the roof vents. – removed to make the report more compact.
Having established a working problem specification by running coarser grid (25x25x15 cells) simulations on PC, some remote MICANET simulations were performed with a finer grid (50x50x30 cells). Figures 15 and 16 show the flow pattern around the warehouse and close-up in a plane through one of the doorways and fire plumes. The vectors are ‘coloured’ by temperature and (smoke) concentration in Figs 15 and 16 respectively. Again, corrupted colours have spoile d the frsges.
The residuals after 1000 iterations were:
Whole-field residuals before solution
with resref values determined by EARTH
& resfac= 1.000E-02
variable resref (res sum)/resref
P1 8.530E-01 1.256E+02
U1 5.503E-01 1.318E+02
V1 7.618E-02 1.042E+03
W1 1.123E-01 2.231E+03
KE 4.128E-02 1.591E+02
EP 6.119E-04 1.628E+03
TEM1 2.203E+04 3.728E+01
PL1 4.933E-04 3.321E+03
Figure 15 shows clearly the wind profile at the inlet and the merging of the external buoyant plume with the boundary layer downstream of the building.
Although the grid is rather coarse, one can observe in Fig 16 re-circulation in the wake of the building. As MICA matures, and automatic grid adaptation is made available, it should be possible to resolve regions of complex flow as they ‘emerge’ from the solution.
Figure 15 Flow pattern around warehouse
Figure 16 Flow in at doorway and rising (internal) plume
6.3 Environmental office
Figures 17 and 18 show some results from a remote simulation of the environmental office test case. Although the solution oscillated (information relayed from CHAM), the overall results are qualitatively correct. The simulation performed 1000 iterations, with false time-steps of 1.0 (momentum equations) and 10.0 (turbulence equations).
The residuals after the 1000 iterations were:
Whole-field residuals before solution
with resref values determined by EARTH
& resfac= 1.000E-02
variable resref (res sum)/resref
P1 8.366E-05 2.831E+04
U1 3.396E-06 4.009E+05
V1 3.622E-06 1.395E+05
W1 6.928E-06 4.583E+05
KE 3.641E-04 1.457E+02
EP 5.346E-04 1.159E+01
TEM1 3.783E+00 1.373E+02
Figure 17 Flow pattern inside office, in ducts and up stack
A comparison simulation was made with the BRE-FRS model JASMINE. The same boundary conditions and source terms were employed. Figure 19 shows the temperature predictions from JASMINE on the same plane as for the MICA frsge. The temperatures generated by MICA and JASMINE are similar.
Figure 18 Temperatures inside office
Figure 19 JASMINE prediction of temperatures inside office
However, a repeat remote simulation was then performed using lower false time-steps, i.e. greater relaxation. Again, 1000 iterations were performed The residuals were lower, as shown below
Whole-field residuals before solution
with resref values determined by EARTH
& resfac= 1.000E-02
variable resref (res sum)/resref
P1 5.874E-05 3.109E+03
U1 1.503E-06 3.722E+05
V1 2.865E-06 7.727E+04
W1 2.781E-05 3.469E+04
KE 3.640E-04 8.969E+00
EP 5.346E-04 9.703E-01
TEM1 2.618E+00 4.683E+01
However this time the solution was different, with flow out of two of the ducts. This resulted in higher temperatures inside the office, as shown in Figure 20. Despite the colour corruption, one can still observe that the temperature inside the office has increased significantly above ambient.
Figure 20 Temperatures inside office using different numerical
relaxation
This highlights the point that auto-relaxation and adaptive grid refinement is required for remote MICA simulations. And furthermore, unless the end user is competent enough to be able to judge the overall convergence of a CFD solution, some expert intervention will most likely be required.
6.4 Forced ventilated nuclear cell
The final test case described here is the forced ventilated nuclear cell. It was not possible to generate the correct solution. The primary reason for this is the inability of SAMSON currently to model high (radiative) boundary heat losses.
Simulations were performed using both fully participating solids (where heat conduction through the solid is modelled) and isothermal boundaries. For the latter, the temperature of the solid was fixed at 150ºC, the approximate value measured in the original experiment. In both cases the temperatures inside the test cell are far too high. This is indicated in Table 1 below, where the temperature of the exhaust gas (at the extract fan) is given. The MICA simulations were performed locally o n PC with fairly coarse grids, but this was sufficient to get the qualitative behaviour.
| Simulation/experiment | Temperature at extract fan (ºC) |
| Experiment | 275 |
| MICA – participating solids | ~1400 |
| MICA – isothermal (150ºC) solids | ~1200 |
| JASMINE | 295 |
Table 1 Temperatures in the forced ventilated test cell
The JASMINE simulation was performed with the six-flux radiation model, a simple solid phase conduction model and a fire/combustion model. A second JASMINE simulation was performed with a heat source instead of a fire source & combustion, which yielded similar results. This indicates strongly that it is the absence of a radiation model that is the cause of the problem in the MICA simulation for this case.
7 MICANET in the context of an industrial environment
The MICA project has demonstrated the potential benefit to engineers working in the building services and fire protection industries, of coupling an ‘easy-to-use’ CFD pre/post-processor with high performance computing resources located at remote (Internet) sites. As stressed in the Introduction, these industries have been slow to take up advanced computer modelling techniques such as CFD. There are, however, many benefits to be gained from making much greater use of CFD for general buildi ng flow calculations, and in particular for smoke control and fire response calculations.
By removing the requirement that the design engineer has an in-depth understanding of the details of a CFD code, and allowing them also to take advantage of high performance computing resources, it should be possible for novel, often more cost-effective, solutions to be generated. This is very pertinent in the fire-protection industry, where the emergence of ever more complicated buildings together with new functional based building codes, has made the need for more sophisticated computer modelling apparent.
Traditional design methods, based often on simplistic empirical calculations, result in many instances in an over-prescribed solution for a smoke control system or structural fire protection specification. If a CFD solution can demonstrate that a more conservative solution can still achieve the required level of fire protection, then the savings made can be very great. A further benefit of using CFD models in the building design stage is in examining a series of alternative fire protection criteria, and then selecting the one most appropriate to the client. In this way novel solutions may emerge.
It is in being able to take advantage of functional based building codes, and to exploit fully the emerging subject of fire safety engineering, that MICANET (or equivalent) offers the most benefit to the fire/smoke control sector. There are both economic and life safety aspects to this. In economic terms, the ability to reduce over-engineering and to allow new novel buildings to satisfy the regulatory authorities will undoubtedly improve the position of the both the construction and fire protection industries. With respect to the
However, as indicated in the Final Recommendations section below, there are still issues to be addressed with MICANET, and this approach in general:
8 Final recommendations
A concise summary, and main recommendations, of the MICA validation work performed by BRE-FRS, form the viewpoint of potential industrial end users in the fire safety engineering community, follows:
