Minutes of the Meeting on
15th April 2002
Present: Keith
Austin (Flowmaster), Sebastien Caillaud (EDF), Christina Giannoppa
(King’s College), Chris Greenshields (Nabla), Martin Hamilton
(Hamilton Flowservices) Thorsten Neuhaus (UMSICHT), George Papadakis
(King’s College), Arris Tijsseling (Eindhoven University), Della
Leslie and Alan Vardy (Dundee University).
Chairman:
Keith Austin
Minutes:
Della Leslie
Meeting
commenced at 09.15 hours.
Apologies: Anton
Bergant (Litostroj E.I. d.o.o.), Jim Brown (Dundee University),
Bruno Brunone (Perugia University), Pierre Moussou (EDF) Simon Pugh
(ESDU), Patrick Vaugrante (EDF), Lixiang Zhang (Kunming University
of Science and Technology).
Preliminary Items
- The minutes from the last meeting were approved.
- New members to the group were welcomed and
participants introduced themselves.
Item
1.
Progress at Dundee – Objectives 1 and 2
The
first two objectives of the project, exactly as specified in the
research proposal, are:
Objective 1: To identify the features of pipe systems that cause greatest
susceptibility to risk of damage through fluid-structure
interaction.
Objective 2: To identify the minimum acceptable capabilities of methods of
analysis suitable for assessing fluid-structure interaction.
Della
Leslie gave a presentation outlining the progress of the current
project. It was emphasised that this was the last Group Meeting that
could strongly influence the technical direction of the current
project, which has only about 6 man months to run. The next meeting
will focus on the content and presentation of guidelines.
The most recent
progress in the project includes development of the ALT in-house
software (Axial/lateral/torsional FSI software for the time-domain).
This is currently in working order, although development will be
on-going as necessary (for the addition of extra features). It has
been used for the first set of parameter variations of System 1
(single pipe). There has also been progress in the experimental
program (details given in second presentation). Della reported that
the FSI web-site is now on-line and that it is intended that this
will be updated regularly and emphasised the need for feedback..
Address: www.dundee.ac.uk/civileng/FSI
Della described the
overall approach to the problem. A large piping system is too
complex to analyse, therefore it is considered as a collection of
sub-systems. Five sub-systems have been identified for use within
the project (as defined in previous meetings). Della pointed out
that system 4, the 3-D elbow pair, was of particular interest
because this configuration is recommended as an expansion group
(although it is commonly found to promote excessive vibration).
For Objective 1, six
generic issues have been identified. These are:
(1)
Long lengths of unsupported or poorly supported pipe work
(2)
Unsupported/unrestrained elbows
(3)
Unsupported/unrestrained valves
(4)
Unsupported/unrestrained or poorly supported T-junctions
(5)
Combinations of the above
(6)
Vibrating machinery
Please see the
web-site for details on each of these features. How each of these
features will be investigated (using sub-systems 1-5) was described.
The aim was to limit the quantity of work required to a manageable
amount, with each generic issue being investigate through only one
or two sub-systems.
Of particular
interest was item (5) T-junctions. A specific example for this is
the EDF problem of failure of small-bore pipes (branching from a
larger pipe). Sebastien added that EDF is currently trying to find
robust configuration (not found yet) and that it is not an easy
problem. Arris said support of T-junction is explicitly mentioned in
the ASME B31 code.
Next Della described
some of the analysis being performed. For each sub-system there are
millions of possible cases. For system 1 (single pipe) the first set
of variations has been carried out and the process of reducing the
number performed was described. We reduce to manageable number by
specifying a finite number of boundary conditions (Closed end,
imposed pressure, semi-infinite, with(out) support, with(out) added
mass). Only certain combinations are realistic for study, e.g.
semi-infinite at left-hand boundary and semi-infinite at right-hand
boundary is not realistic, also the combination of a rigid support
with an added mass is not realistic. Della illustrated the axial
impact of a single pipe example, showing that the 486 possible
combinations of boundary conditions in fact reduce to 96. For each
example the output parameters are hoop stress (and strain), axial
stress (and strain), Tresca stress (and strain) axial and lateral
displacements and pressure. The maximum value at each location along
the pipe system is recorded. Some examples of results were shown and
discussed (the results will be displayed on the web-site)
For Objective 2, the
assessment of methods, there are many methods that may be
considered. Della identified those methods that are being
investigated and particular features of each of these. Briefly these
were: Joukowsky (basic, simple guide: many
cases sufficient); Waterhammer (good for rigid systems and buried
systems, but misses structural modes of response); Structural
dynamics (misses fluid response). For FSI we have: FSI: Friction
Coupling (Significant for very long pipes. Though not expected to
contribute to undesirable behaviour); FSI: Poisson Coupling (Highly
dependent on material properties (e.g. flexible pipes)); FSI:
Junction Coupling (Most significant type. Interactions at junctions
can cause significant change in measures pressures/stresses.)
Finally uncoupled FSI was mentioned, although investigation of this
will not be carried out (possible future work). We intend to
acknowledge that it is common practice to use this method and it is
expect a significant number of cases where this method is sufficient
(e.g. will predict the magnitude of forces.) There was a discussion
on uncoupled FSI and its use.
Item
2.
Low-frequency vibrations of piping
Sebastien
Caillaud gave a presentation on the work of his colleague Pierre
Moussou. This continues from work presented at the previous meeting
(October 2001).
The
research is based on changing the method of assessment. Previously a
numerical approach was used (Circus), but this has been found not
the answer. The new method based on experimental results (and
simulations). Though it was noted that it is expensive to take
measurements in Nuclear Power plants. Starting from the basics, the
design of a system is standard. The input data is the pressure class
and the nominal flow. Pipe cost vs. pumping cost determined pipes
diameter; pressure class and diameter determines pipe thickness and
weight conception determines support spacing. Experience shows that
natural frequencies fit into two ranges 5-30Hz and 50-200Hz. This is
a generic result (has been found in several systems). Sebastien
showed that the natural frequencies (analytical) on a single span
beam with pin-pin condition gives two ranges, 0-50Hz and 50-200Hz,
as function of distance between supports. The results found were
discussed and the definition of the parameter alpha, which was part
of the formula giving support spacing was requested:
alpha
= sqrt(density_steel*momentum_inertia_pipe/(lambda*outer_diameter2)
)
where
lambda* length_pipe = mass of the pipe
which
gives 0.2 < alpha < 0.3 for the pipes studied.
The
overall idea is to analyse the general behaviour of a pipe system.
Sebastien showed show graphs giving the frequency response under
nominal and partial regimes (measured on-site). On each figure, the
general shape for each regime was given, proving an envelope for
acceptable operation of a system. The figures suggested a
logarithmic threshold for ‘normal spectra’.
The
remainder of the presentation looked at hydroacoustic sources in
systems (including work by Gibert). He also presented pressure
spectra according to literature, but said that there was a lack of
accurate data for industrial use.
Item
3.
WAHALoads project
- two-phase flow for waterhammer and induced loads
Thorsten
Neuhaus gave a presentation on the WAHALoads project; task T3.3
–Benchmark Calculations with existing codes. There are seven
participants, eight codes involved and three benchmark problems. He
described each code being tested: CATHERE V1.5a (CEA); RELAP5/mod3.2
(EA, TBL); EUROPLEXUS 2000 (EDF); WAHA single phase code (IJS); TFTC
Two-phase Flow test code (IJS); DELOS (UCL); MONA 2.2 (UMSICHT).
Alan suggested it would be useful if different institution used the
same code for the same problem (to see if they got the same results).
Summarising, there are two types
of codes, general-purpose system codes and specialised fast
transient codes.
The
benchmark problems are outlining below together with some of the
conclusions.
BM1.1a:
shock tube theoretical test case – cold liquid, single phase. A
compression wave is generated by sudden valve closure at pipe
downstream end. The lower pressure remains above saturation
temperature and no vapour is generated. It was found that most codes
were capable of producing acceptable results, provided sufficient
nodes and sufficiently small time steps are used.
BM1.1b:
shock tube theoretical test case – hot liquid, two-phase
conditions. A compression wave is generated by sudden valve closure
at pipe downstream end. The pressure drops to saturation
temperature, so that vapour generation and condensation can occur,
leading to possible condensation shocks. General-purpose codes were
not capable of capturing secondary effect due to vapour cavity
collapse, but the special purpose codes for fast dynamic transients
could. Thorsten played some simulations showing the results provided
by the codes.
BM2:
Edwards Pipe Blowdown experiment. The pipe is initially filled with
pressurised, subcooled water and the transient initiated by the
burst of the rupture disk located at one end of the pipe. A
rarefaction wave travels towards the closed end of the pipe and
consecutive blowdown is controlled by flashing. Each code predicted
the rarefaction wave, but there was a problem with flashing model in
MONA code. All codes predicted a higher pressure during first phase.
It was suggested that the boundary condition might not be formulated
correctly, since all code gave the same result.
BM3:
PPP tests (power plant pipe work). This system was described and the
initiating event was a rapid valve closure. Thorsten presented the
results given by MONA and also Flowmaster. The results were
discussed. The timing of the pressure pulses were well predicted by
MONA, but the magnitudes were incorrect.
During
the presentation there was extensive discussion on the definition of
bubble flow, column flow and other definitions of two-phase flows
according to the value of the void fraction.
Item
4.
Development of Guidelines – Objective 3
To express these outcomes in a manner that will reduce significantly the
uncertainties faced by designers of pipe systems.
Della began by first presentation progress of
the experimental program at Dundee. Last year experiments on the LT
system (in-plane elbow and T-piece combination) were carried out for
an air filled system, with the help and guidance of Arris
Tijsseling. There had been several problems with leaks that had
prevented the water filled experiments being performed. In January
2002, the experimental work was continued (now working without
external assistance, but with the help of an excellent technician
who has assisted previously on the system – Colin Stark). Previous
problem had been solved, though there were still many new one. These
included some new leaks (at a pressure transducer mounting, at the
T-junction and a bleed screw). More important was the incorrect
fitting of the o-ring in the T-junction, which was discovered only
after high-velocity impact testing lead to a leak. Other problems
involved updating of equipment and associated incompatibilities.
Over 50 test were completed (125kHz and 10KHz sampling rates, with
high and low-velocity axial impact and a lateral hammer impact)
taking pressure and strain reading at several locations along the
system.
Della illustrated
some of the results obtained. Consistency in the results was shown
to be excellent. She continued by comparing the experimental results
with numerical simulations (using in-house software). These appear
to give good agreement, although the numerical simulation exhibited
some high-frequency oscillations. It was suggested that this may be
due to the extremely fine grid used in the numerical work (time step
was of the order 10-7 seconds.
The
main part of this presentation was the discussion of Objective 3:
How to express outcomes of the project. Della gave an overview of
the FSI Web site that has been developed, and will continue to be
developed. The site will used for displaying results and act as
focus for FSI research. As results and data is obtained they will be
put on the site, allowing for immediate feedback.
The
development of the site was discussed, including ideas to for items
and links on the site. Arris suggested links to Dave Wiggert’s
professional course (on FSI) and also items like previous EPSRC
reports. Della said that contributions to the site would be
welcomed, in particular anyone with access to photographs suitable
for the site.
For
the development of guidelines, Della described the ideal solution to
the problem as a tree-type diagram enabling engineers to determine
whether a system would a susceptible to FSI. It was emphasised that
the current project will achieve only a small part of this; the aim
of the project being to reduce uncertainties, not eliminate them.
Alan said that the results should answer the following: 5 subsystems
as defined – are these in your system? Will they cause a problem?
It is important to keep results found simple and make them
accessible to engineers.
Finally
the discussion moved on to the future of research at Dundee. The
current project finishes this year and Della will be moving to
Ireland. Arris asked if the experimental apparatus would survive?
Yes (hopefully)! Alan said that future plans would be made once the
current project was closer to completion.
Item
5.
Casto1: Experiments and calculations
Sebastien
Caillaud gave a presentation, reporting the experimental and
numerical analysis of Castro1: a single elbow system. The system was
designed to focus on coupling in the elbow. The system is supported
rigidly at one end and is open at the other. Experiments have been
carried out with the system empty, i.e. air filled and water filled
(the level of the water is varied). The system is excited (using an
oscillator) at the remote (open) end, in-plane with the system.
For
the air-filed system, experiments were performed and the results
compare to simulations by both Circus and Code_Aster. The initial
model (using Circus) yielded errors of about 3% and an updated model
reduced this to about 1.5%.
For
the water filled system there had been many problems, particularly
with air in the system. Sebastien said that the system was filled
with water, then shook and left for several months before
measurements were taken. There was a discussion of other possible
solutions for this problem.
Sebastien presented
the results for the water filled system (error level was again about
3%). He showed comparisons between experimental results and
different numerical models: Circus coupled, Circus with added mass
and Circus acoustic (i.e. comparing FSI and uncoupled model). It
showed that uncoupled results and coupled method were very close,
particularly for first two modes (higher modes show a greater
difference). This implies that the effects of added mass are
dominant and coupling effects in the elbow are weak.
Sebastien continued
by outlining the future work at EDF. This includes Casto2, which is
a Z-shaped pipe (2-elbows in plane). The experiments for the air
filled system have been completed and the experiments for the
water-filled system will be coming soon.
Item
6.
Progress in FSI in flexible tubes
Christina
Giannoppa gave a presentation reporting the progress in her PhD
project on FSI in flexible tubes. She began by giving an overview of
the project and the approach used for modelling: a single solution
method using a finite volume code (Foam). She described how the
equations were derived. This requires the equations for the solid
being written in terms of variables traditionally used for fluids
(Velocity and pressure). Because this is a new formulation the code
needs to be validated for solids and the test case for this is a
beam, fixed at one end and subjected to a step load at the remote
end.
Typical
stress analysis (displacement based) shows dissipation (which
shouldn’t occur). Using a velocity-based form with the pressure
calculated explicitly, two methods were examined: Euler implicit and
backward difference. Both showed less damping, but the backward
difference method exhibited oscillations. Question whether this was
physical or numerical. Repeating calculations with various time
steps altered the frequency of the beat. The group discussed this
and it was concluded that this must be numerical. Arris suggested
using a forward-difference method. Comparing results between the
numerical (3-D) and analytical (1-D) showed a difference of 8.82%
for mean and maximum displacement and a 0.9% for the frequency. The
group discussed whether comparisons should be made between 1-D and
2-D models, emphasising that comparison must be expressed as %
differences and not % error.
Christina
continued by describing the method with pressure included
implicitly. For the formulation of the boundary conditions there
were two options: velocity with a fixed gradient or a fixed value,
together with constant pressure. Depending on which is used gave
different results, one gives energy loss and the other energy gain.
The reason for this was unknown, as both should provide the same
result. This was discussed and it was pointed out that the case
under consideration was very demanding. It was suggested that a pipe
beam might be a useful example to investigate, particularly since
the next step would be a flexible tube. For the formulation of a
flexible tube it was highlighted that pressure would be
discontinuous across the solid/fluid interface. Possible methods to
overcome this were discussed.
Any Other
Business: A paper by Lixiang Zhang was distributed.
Next
meeting: October 7th 2002
Item
7.
Chairman’s Closure
Closure
of meeting: at 17:30 approx.
|