Professional Documents
Culture Documents
MT25 Paper TFCC Simulation v4.0 FINAL VERSION
MT25 Paper TFCC Simulation v4.0 FINAL VERSION
MT25 Paper TFCC Simulation v4.0 FINAL VERSION
09-10 1
The work leading to this publication has been funded by Fusion for Energy
under the contract F4E-OPE-414. This publication reflects the views only of
the author, and Fusion for Energy cannot be held responsible for any use
which may be made of the information contained therein.
The views and opinions expressed herein do not necessarily reflect those of
the ITER Organization.
M. Jimenez is with Fusion For energy, Barcelona, Spain, Europe
(email:marc.jimenez@f4e.europa.eu)
Fig. 1. TFC overall dimensions in mm and TFCC main sub-assemblies
Template version 8.0, 27 July 2017. IEEE will put copyright information in this area
See http://www.ieee.org/publications_standards/publications/rights/indes.html for more information.
2
Welding operation is delicate, since the WP should be pro- complexity was increased to scale 1:1 sections of the real
tected from the generated heat, and the displacement of the TFCC but limited to 1m in length. In this way, the interaction
cases due to the welding distortions should be as limited as of the different welding passes and different geometries were
possible. As example, 0.5 mm in planarity tolerance has to be tested and adapted to the FEM tool. With this, the FEM tool
respected in the ~10 m of straight section of the AU case. Pre- was ready to be used in the full TFCC geometry in order to
vious studies [2] showed that welding thick SS structures cre- predict deformations of the real case. A scheme of this meth-
ate non-negligible distortions which could cause difficulties to odology is presented in Figure 3.
meet dimensional tolerances. As mentioned previously, the The second aspect of the strategy was to simplify the nu-
TFCs are components with numerous interfaces, so defor- meric problem to be solved as much as possible, in order to
mation and out-of-tolerance interfacing areas are to be avoided obtain reasonable computing time, while not compromising
to ensure that each TFC can be assembled properly. In this re- too much the accuracy level.
gard, extra-material or over-metal is present in the interfacing
area, which should be machined once the welding is finished.
Since this is a costly and time consuming operation the
amount of extra-material has to be minimized. This complex
scenario leads to the need of a strategy to assess welding dis-
tortions in an efficient and accurate manner. In order to face
this, a campaign consisting in real welding of TFCC-like
structures was launched in order to feed experimental data to
create a suitable FEM tool. Typically, welding simulation re-
quires coupled thermo-mechanical transient models to be
solved, which require a large amount of time for industry re-
lated complicated geometries. For this reason, several simpli-
fication techniques are typically used, such as mesh condensa-
tion techniques [2], or imposed inherent strain [3].
This paper presents the strategy followed by EnginSoft, F4E Fig. 3. Global methodology flowchart. Experimental activities on welding
and SIMIC for the development of the FEM tool, starting with coupons and mockups are used to define and fine tune the FEM tool to be
an overview of the methodology, results assessment, conclu- used with the full TFCC geometry.
sion and further work.
For this, three main decisions were made: First, the applica-
tion of equivalent space-time temperature maps to mechanical
models, thus de-coupling the thermal-transient problem, using
a quasi-steady state analytical solution for a moving heat
source following Rosenthal [5] and others posterior studies
[6], [7], for thick plates. Second, a mesh reduction via cluster-
ing weld passes, as per [2]: this idea consists in the condensa-
tion of several passes within a weld, maintaining as much as
possible the real sequence in the deposit of the passes. For
this, the history of displacements predicted should be matched
with the experimental data. Third, to use techniques developed
Fig. 2. Weld chamfers for manual (right) and automatic (left) welding. for laser cut simulation in a reverse way, using ANSYS ®
technology of element ‘birth’ and ‘death’ studies [8]. The pro-
cedures have been developed by applying thermal power den-
II. METHODOLOGY sity on each element under the action of the electrical arc as
A. Global Approach body force, while the heat transfer is a function of the arc dis-
placement along the welding. For each pass, a layer of ele-
The strategy adopted for the execution of the activity was ments is set active row by row as the arc moves along it. Heat
based in three main aspects. The first aspect was the prepara- is conducted through the bottom and side nodal surfaces of the
tion and execution of a set of experimental welding exercises. forming chord, while it is radiated through the top one through
These welded structures had to be representative of the final the use of surface radiation elements which are kept “alive”
TFCC geometry but, at the same time, simple at the beginning, until the following pass occurs and then killed. Until the next
increasing their complexity in further steps in order to be used row of elements is activated, the front nodal surface is kept ad-
as an experimental feeding system to the FEM tool to be de- iabatic. All these numeric techniques were used to feed the
veloped. The initial welding exercises were the welding cou- FEM tool via a series of tuning parameters: cut-off tempera-
pons used for welding qualification [4]. They were used to es- ture, i.e. temperature at which an element is activated, chord
tablish the main aspects of the FEM tool, with one of the cou- mesh size and coefficient of thermal expansion (CTE) of
pons being used as a blind test. Then, the welding structures chord material.
3
The third aspect of the strategy was the need of obtaining reli- C. Activities on true TFCC section mockups
able experimental results, and how to treat and process them in As next step, three full scale 1 m long mockups were weld-
order to be used to feed the FEM tool. Usually, this kind of ed and simulated, in order to validate and fine tune, if neces-
simplified FEM weld simulation and tuning exercises are exe- sary, the calibration parameters of the FEM model, choosing
cuted at laboratory scale. Indeed, for coupon outputs, it was the CTE of the chord material as main calibration parameter,
simple to instrument and obtain displacement and temperature as a first choice. The mockups were named, respectively,
data. On the other hand, scale 1:1 mockups were big, massive mockup A (MU-A), representing an equatorial section of the
structures composed of very thick, up to 120mm, steel plates, AU TFCC part, mockup C (MU-C) representing an equatorial
which should be welded and handled in an industrial work- section of BU TFCC part and mockup D (MU-D), representa-
shop environment. For this, an extensive metrology work was tive of the AU-BU upper connection zone.
necessary in order to obtain reliable data depicting the defor- The welding of the three mockups required several months,
mation of the structures for each welding step, in order to use during which position of the control points was monitored ap-
this data to feed the FEM tool. proximately every 25mm of deposited welded material. Re-
More details of each phase of the activity are given in the fol- garding FE modeling, the FEM procedure developed on cou-
lowing sections. pons allowed to successfully carry out the all the welding sim-
ulations. The condensation of the passes allowed keeping the
B. Welding qualification coupons computational time reasonably low, even for MU-D, with the
highest number of welding chords, whose solution time took
A total of six qualification coupons were welded, repre-
less than 1 week.
sentative of the characteristic welds to be executed on the real
TFCC. Most coupons (5 out of 6) were manually GTAW
welded, while 1 coupon was executed to qualify the automatic
welding of the TFCC. Coupons’ thicknesses were 60mm and
65 mm, lengths 850mm to 900mm and width 300mm and 500
mm. Vertical and horizontal displacements and temperature
measured by thermo-couples in the coupons were used to feed
the simulation and calibration activities. One of the coupons
was used as blind test, i.e. was welded after the simulation re-
sults were obtained, in order to avoid any bias when compar-
ing the obtained and expected results. Figure 4 shows the out-
come of the blind test, where the experimental vertical dis-
placement of the free plate was compared to the simulation re-
sults and a prior estimation of considering the displacement as
a cubic function in time As a consequence, the CTE and mesh
sizing of the chord material were defined, and also all the load
inputs and cooling laws scripts which would define the FEM Fig. 5. Drawing of MU-D with control points and butt weld – red arrows -
tool to be used in the upcoming mockup simulations. representation. Distance variation caused by the welding process of the close
points along the weld, such as the circled pair, is represented the figure 6.
MU-D presented another difficulty as well; the custom ma-
chining of the splice plate, a piece which will ‘close’ the
mockup once the AU-BU and AP-BP parts are welded. This
custom machining was required to be replicated in the FEM
model, which required a challenging adaptation of the well-
known ANSYS® sub-modeling technique – imposed dis-
placement at a cut boundary- so that the splice plate would fit
into the already deformed structure due to the welding distor-
tion caused by the precedent welding passes. In addition, a
procedure for the numerical-experimental comparison was
successfully created; such technique is based on the compari-
son of variations of distance between control points, located
both in the real part, obtained after metrology surveys with
Laser tracker, and also in the FEM models, assessing thus the
distortion of the structure caused by the welding. The mockup
Fig. 4 Plots of coupon vertical displacement during welding process, used for
study and the analysis activity, by comparing the deformation
blind test and after final tuning of the FEM tool.
results predicted with the measured ones, indicated the need of
increasing the CTE value, as calibration parameter, for the au-
tomatic welding of MU-A and MU-C +34%, to take into ac-
4
REFERENCES
[1] N. Mitchell, et Al, The ITER Magnet System, IEEE Trans. On Applied
Superconductivity, pp. 435-440, 2008.
[2] J. Caixas et Al, Weld Distortion Prediction and Control of the ITER
Vacuum Vessel using Finite Element simulations, Fusion Engineering
and Design, Volume 88, Issues 9–10, Pages 2011-2014, 2013
[3] H. Murakawa et Al, Applications of inherent strain and interface ele-
ment to simulation of welding deformation in thin plate structures.
Computational Materials Science. 51. 43–52. 2012
[4] B. Bellesia et Al, Progress on European ITER TF Coil Cold Test and In-
sertion Contract. IEEE Transactions on Applied Superconductivity. 26.
1-1. 2016
[5] D. Rosenthal, The theory of moving sources of heat and its application
to metal treatments. Transactions of the American Society of Mechani-
cal Engineers 68:849- 866, 1946
[6] D. Darmadi et Al, Analytic and Finite Element Solutions for Tempera-
ture Profiles in Welding using Varied Heat Source Models. World
Academy of Science, Engineering and Technology, 81,154-162, 2011
[7] A. Ghosh, Analytical Modeling of Moving Heat Source for GMAW, In-
ternational Journal of Mechanical Engineering and Research, ISSN
No.2249-0019, Volume 3, Number 4, pp.335-344, 2013
[8] A. Capriccioli et Al, Multipurpose ANSYS FE procedure for welding
processes simulation, Fusion Engineering and Design, Volume 84, Is-
sues 2–6, Pages 546–553, 2009