You are on page 1of 13

SIMULATION OF COMPLEX NON-LINEAR STRUCTURES IN LARGE SCALE ANALYSIS

SIMULATION OF COMPLEX NON-LINEAR STRUCTURES IN LARGE SCALE ANALYSIS

N. Jaksic (*), J. Simon-Weidner (*), P. v. Eeten (*), M. Sochor (*) (*)Engineer/Scientist, Max-Planck-Institute for Plasma Physics, Germany THEME Confidence in Engineering Analysis Results KEYWORDS Structural Analysis, Benchmark, Superconducting Coils, Fusion, Stellarator. SUMMARY A large steady-state plasma fusion experimental device of the stellarator family is being developed at Max-Planck-Institute for Plasma Physics (IPP), branch institute Greifswald Germany. The ultimate aim of fusion development (R&D) work is to make fusion energy usable for energy supply, mainly for providing electrical energy. The fundamental principle of the energy extraction by fusion is based on fusion of light atoms (e.g. hydrogen isotopes) in a very hot gas (plasma) which is confined by a strong magnetic field. The magnetic field is generated by superconducting coils which produce the magnetic induction of 3T at the centre of the plasma column [3]. In order to ensure the precise localisation of the magnetic field confining the plasma, the central pathway of each of the individually shaped coils should be predictable during operation with high accuracy. In relation to the size of the complete experiment of roughly 15 m (total diameter), the required position accuracy of the main 4 components is in the range of 0.5- 1.5 10 . An important characteristic of the magnet coil structure is the multiple interior symmetry conditioned by physically demands. Additionally to the positioning tolerances, angle tolerances of 0.5 are prescribed to reduce the deviation of axial symmetry. The superconducting magnets have to be kept at cryogenic temperature (4K). Next important characteristics of the experimental device are the extremely high temperature differences within the structural elements which are partly very close to each other. Some parts of the plasma surrounding structure are loaded by a power of 10 MW / m 2 locally. E.g. it is a big physically and

SIMULATION OF COMPLEX NON-LINEAR STRUCTURES IN LARGE SCALE ANALYSIS

technically challenge to design a thermal insulation for a device with central temperatures in the range of 100 mill. K (plasma column centre) and 4 K (superconducting coil) with a distance of about 0.6m (Fig. 2). By reason of the small distance between the structural elements (< 10 mm) with a temperature difference of about 300 K, it is important to avoid a contact between the elements during operation. A contact of structural elements with different temperatures causes high cryogenic losses and could initiate a so called quench (loss of the superconducting properties) of the superconducting coils. Therefore, it is necessary to predict the element position during operation as precise as possible. The behaviour of such a complex structure can be reliable predicted by means of extensive finite element (FE) analyses only. The principal aim of structural analysis is to create a global model of the experimental device and additionally a majority of sub-models for each system and critical components. The global model has been extensively used for mechanical and thermal analyses [2]. These analyses have shown, that the structure reacts highly sensitive onto changes of initial contact gap widths, contact friction factor and other parameters like bolt pretension, general coil stiffness definition, etc.. The main reasons for the high sensitivity of the structure are multiple internal geometrical nonlinearities. As a consequence, it has been decided to perform a benchmark analysis with an entirely independent FE model with the aim to increase the confidence on the analysis results. The benchmark analysis procedure is the main subject of this paper. 1: Introduction The plasma fusion research is focused world wide on two different types of experimental devices tokamak and stellarator with different principles of plasma confinement. The world's largest plasma fusion experimental device of the stellarator type, called Wendelstein-7X (W7-X) (Fig. 1), is presently being built at the Max-Planck-Institute for Plasmaphysics (IPP). The aim of the experiment is to prove the reactor validity of the stellarator research line in addition to the tokamak research line. The inherent property of the stellarators is a true steady state operation which is of major interest for fusion reactors. The superconducting magnet system with the support structure represents the core of the device and will be the main subject. The important characteristic of the magnet geometry is the five-fold symmetry and the modularity of the coil arrangement system (Fig. 4). The coil system consists of 50 module field coils (MF) and 20 ancillary field coils (AF). Due to symmetry conditions within a module, there exist five MF and two AF coils types only. The coils are arranged toroidally inside a cryostat which has an average large diameter of

SIMULATION OF COMPLEX NON-LINEAR STRUCTURES IN LARGE SCALE ANALYSIS

11m. The MF coils are wound with 6, the AF coils with 3 double layers of a cable in conduit conductor, which is cooled by forced flow supercritical helium at a temperature of approximately 4 K.

Figure 1: Figure 2: the

Assembly of the fusion experimental device W7-X, a birds eye view CrossSection through

outer vessel
experimental device, = 0 0

SIMULATION OF COMPLEX NON-LINEAR STRUCTURES IN LARGE SCALE ANALYSIS

Figure 3:

Assembly of a superconducting magnet coil W7-X

The cable is build by 243 NbiTi/Cu strands which are enclosed by an aluminium alloy jacket (16 x 16 mm), shown in Figure 3. The winding pack by itself is not strong enough to withstand the high electromagnetic forces [4]. Therefore, individually adapted coil housings (e.g. the cast steel 1.3960) with sufficient cross section have to be designed to encapsulate the windings and to support the remaining integral forces. The embedding of a quartz sand-armed resin guarantees a gap-less transmission of the electromagnetic forces of the superconducting strands through the conductor jackets via the basic insulation to the coil housing. The considerable residual forces of the modules have to be balanced: First by a central support ring, secondly by lateral support elements between the coils and thirdly by additional lateral pads near the inside circumference of the coil set. The main features of the superconducting coil system are summarized in Table 1.
Table 1: W7-X - coil system parameters Number of module field coils (MF) Number of ancillary coils (AF) Main radius of module field coils Min. distance plasma-coil Max. Current in the conductor Max. Magnetic induction on axis Max. Magnetic induction at the coil Max. Stored magnetic field energy Time for discharge of magnetic energy Max. Force on a coil 50 20 1.5 m 0.3 m 18.2 kA 3.0 T 6.7 T 0.62 GJ 5s 3.76 MN

current

He in/out

2: Description of the FE-Model 2.1: General The FE-model is reduced to 1/10th of the whole system by the introduction of special boundary conditions and consists of 5 MF and 2 AF coils (Fig. 5). This is possible because the original geometry as well as loading of the structure

SIMULATION OF COMPLEX NON-LINEAR STRUCTURES IN LARGE SCALE ANALYSIS

obey specific symmetry properties. The derivation of the special symmetry definition is described detailed in [1]. Neglecting the influence of gravity, a 36 sector model can be built using these boundary conditions. Considering the gravity, an extension to a 72 sector model has to be realized.

Figure 4:

Coil arrangement of fusion experimental device with axes of symmetry

2.2: FE-Model Preparation The FE-Model for the benchmark analysis has been done very carefully taking into account all essential physically and geometrically features of the structure as accurate as possible keeping in mind the global model size. Here some examples: Modelling of geometrical nonlinearities contact surfaces Modelling of structural junction elements bolts, weld seams Modelling of special shaped contact surfaces spherical interface boundary surface Taking into account different material and frictional properties

SIMULATION OF COMPLEX NON-LINEAR STRUCTURES IN LARGE SCALE ANALYSIS

Figure 5:

FE-Model of fusion experimental device top view

Figure 6:

Suspension of the coil casing 1 on the support ring side view to the coil

SIMULATION OF COMPLEX NON-LINEAR STRUCTURES IN LARGE SCALE ANALYSIS

a.)
Figure 7:

b.)

Discretization of the coil cross-section a.) MF-coil and b.) AF-coil

Figure 8:

Narrow Support Element (Cross-Section)

Figure 9:

Central Support Extension (Cross-Section)

Table 2: FE-Model main parameters Node number Element number Number of contact surfaces pairs Number of contact surfaces Number of contact segments 368369 340000 11 104 100000

The general principle of the coil support structure (Fig. 5) is: Each particular coil housing is fully separated from each other and has to carry the primary load caused by electromagnetic forces due to the current of the coils within the magnetic field. The toroidal arrangement of the MF coil housing is fixed to the central support ring with special joints, called central support elements (CSE), at two locations (Fig. 6) only. Additionally, a mutual support between adjacent coil housings is realized by intermediate joints which are manly loaded by compressive loads. These intermediate joints (located at different positions at the circumference of the coils) are operating with different functions: At the regions of small distances between the coils (internal segment of the coil

SIMULATION OF COMPLEX NON-LINEAR STRUCTURES IN LARGE SCALE ANALYSIS

housing), the intermediate joints, realized by contact elements, are balancing the forces between two adjacent coils. These intermediate joints, so called narrow support element (NSE), (Fig. 8) have an essential effect on the lateral stiffening of the structure. Other intermediate joints, so called lateral support element (LSE), are responsible for the stabilization of the equilibrium of the coil arrangement in lateral direction. The housings of the AF coils are fixed to the central support ring at two locations in a similar way like MF coils. Additionally, these AF coils are fixed laterally at both sides. This fixation is realized using the contact surface capability so that the radial displacement (viewed in local coordinate system of the coil) is performed without constraints. The counter part of the lateral fixation, called planar support elements (PSE), is constrained to the housings of the MF coils. Additional contact elements (CE) positioned at the top and bottom side of the coils in the half module symmetry surface increase the lateral stiffness of the coil arrangement additionally. The coil housing connection of the both MF and AF coils to the support ring (via CSE) has been realized using a bolted connection (Fig 9). Due to the geometrical requirements, the CSE consist of 2 separated parts, one belongs to the coil housing and the other one to the support ring. This involves some additional investigations of the connecting elements. 3: Benchmark analysis results The ADINA [5, 6] code has been used for the present benchmark analysis whereas the ANSYS [7] code has been used for the long-term analysis. The most important differences in the model definition are summarized in table 3.

Table 3: Main models definition differences

ANSYS
General Mesh Definition inside of Solids Mesh Continuity NSE Geometry Definition Bolt and Pin Distribution (Planar Coils) Solution Contact Algorithm

ADINA

Discontinuous

Continuous

Approximately

CAD-Model complying

Approximately

CAD-Model complying

Penalty Method

Constraint Function

SIMULATION OF COMPLEX NON-LINEAR STRUCTURES IN LARGE SCALE ANALYSIS

Model Mesh Size Totally

360000 Elements

340000 Elements

Because of the necessity to make the results of both models comparable, the general boundary conditions for both FE-Models have been mutually agreed. Correspondingly, the so called Low-Iota load case has been applied to the 36 model, conditions like cool down and gravitational forces are neglected. A first information of the behaviour of the structure for a particular boundary conditions case is given by the result of the displacements (Fig. 10).

Figure 10:

Displacement Magnitude of ADINA Model Top View

The comparison of the displacements given by the ANSYS code shows similar values. Nearly the same displacement pattern could be found within the whole structure in both cases. The maximum displacement (limited on a small quite sensitive area) amounts by the ADINA model 18.66 mm and by the ANSYS model 16.5 mm. The next step in the benchmark analysis belongs to the comparison of the cross-sectional forces and moments for the most stressed structural elements like CSE and LSE. By reason of comparison, the cutting surface for the evaluation of the cross-sectional loads has been exactly predefined.

SIMULATION OF COMPLEX NON-LINEAR STRUCTURES IN LARGE SCALE ANALYSIS

The notation in the table 4 e.g. NPC1Z1 means central support extension at coil 1 upper and NPC1Z2 means central support extension at coil 1 lower, etc..
Table 4: Cross-Sectional Forces at CSE
ANSYS Fx MN 0,714 -3,146 -0,764 -0,651 -1,021 -2,080 0,130 -2,016 0,055 1,446 ADINA Fx MN 0,714 -3,317 -0,896 -0,605 -1,047 -2,088 0,161 -2,022 0,082 1,578 dev % 0,0% 5,2% 9,6% 3,7% 2,2% 0,3% 3,6% 0,3% 3,4% 8,7% ANSYS Fy MN 0,259 0,242 -0,332 -0,002 0,186 0,614 -0,036 0,106 -0,703 -0,097 ADINA Fy MN 0,177 0,189 -0,431 -0,074 0,080 0,661 -0,007 0,014 -0,754 -0,106 dev % 5,9% 1,6% 7,2% 5,8% 8,7% 2,0% 3,4% 4,3% 6,3% 0,6% ANSYS Fz MN 1,214 -0,707 1,015 -1,079 0,617 -0,909 -0,861 -0,746 -0,375 -0,090 ADINA Fz MN 1,082 -0,568 1,050 -1,073 0,635 -0,888 -0,830 -0,841 -0,355 -0,119 Dev % 9,6% 4,2% 2,5% 0,5% 1,4% 0,9% 3,5% 4,4% 2,5% 1,9%

CSE NPC1Z1 NPC1Z2 NPC2Z1 NPC2Z2 NPC3Z1 NPC3Z2 NPC4Z1 NPC4Z2 NPC5Z1 NPC5Z2

Table 5: Cross-Sectional Forces at LSE


ANSYS ADINA ANSYS ADINA ANSYS ADINA Fx Fx dev Fy Fy dev Fz Fz dev MN MN % MN MN % MN MN % 0,305 0,277 2,0% 1,032 0,933 7,0% -0,983 -0,964 1,3% 0,344 0,327 1,9% 0,402 0,342 7,1% 0,665 0,716 6,0% 0,304 0,302 0,5% 0,266 0,233 6,8% 0,277 0,294 3,5% 0,080 0,095 2,1% 0,703 0,686 2,5% -0,043 -0,073 4,2% -0,886 -0,936 3,6% 1,055 1,088 2,3% 0,087 0,068 1,4% -1,058 -1,069 0,8% 1,030 1,082 3,5% 0,134 0,142 0,6%

LSE LSE1-1 LSE1-2 LSE2-3 LSE3-4 LSE4-5 LSE5-6

The notation in the table 5 e.g. LSE1-1 means lateral support element between coil type 1 and 50 and LSE1-2 means lateral support element between coil type 1 and 2, etc.. The comparison of cross-sectional forces and moments is very important and a very strictly criterion too, for the confidence to the results of the benchmark analysis. The result values vary with a slightly position change within the area of interest, the fineness and the style of the mesh might influence the results as well. Nevertheless, the comparison of the crosssectional forces and moments show sufficient agreement. For the comparison, the significant values of the results have been taken only. Values with low contribution have been neglected by reason of relevance. The comparison of the cross-sectional forces (Tables 4 and 5) show a total maximum deviation of less than 10%. The maximum deviation of correspondent cross-sectional moments has been found to be within 30%.

SIMULATION OF COMPLEX NON-LINEAR STRUCTURES IN LARGE SCALE ANALYSIS

Furthermore, the contact forces at the NSEs have been compared as well. Note, the contact property definition in the models have been done by a different way also. Moreover, the FE models use different contact solution algorithms (Table 3) to solve the contact problems.
Table 6: Contact Forces
ANSYS Pads 50E2--> 1E2 Fx MN 0,692 ADINA Fx MN 0,624 dev % 10,2%

In spite of entirely different contact definition, the contact forces are in good agreement as well. The example in table 6 for the contact surfaces between the coils 1 and 50, which are representative for relevant values, shows a contact force deviation of about 10%. A maximum deviation of the contact forces for contact surfaces with fewer relevancies has been found to be within 30%. Additionally to the above described benchmarking, the ADINA global model allows an advanced analysis of all narrow support elements. Such a detailed analysis of all NSEs was not possible in global model of the longterm analysis. This part of the analysis has been done by means of trials for a particular element and a particular load as a worst case analysis. Correspondingly, the model can be used for an additional result evaluation like contact surface pressure, contact surface area in contact, contact slipping, etc.. A confident analysis for narrow support element results is extremely important for the reliability of the whole device since these elements are the most stressed elements in the structure with loads at the mechanical limit. Figures 9 and 10 show an example of contact surface pressure and contact surface sliding for an individual narrow support element between coil housing 1 and 50.

SIMULATION OF COMPLEX NON-LINEAR STRUCTURES IN LARGE SCALE ANALYSIS


Figure 11: Contact Surface Pressure Sliding Figure 12: Contact Surface

The notation in the figures above e.g. NPS1E8-NPS50E9 denotes a contact surface pair between coil casing 1 and 50. 4: Conclusions This paper describes the first step of a large scale benchmark analysis for a complex structure of a fusion experimental device. After additional improvements of the 36 benchmark model the computing will be expanded to a 72 model. Present benchmark analysis has been done by comparison of the results of two FE-Models which have been modelled independently of each other. The finite element analysis has been done by different FE codes as well. In the present case the ANSYS code has been used for the long-term analysis whereas the ADINA code has been used for the present benchmark analysis. The comparison of the results of the benchmark analysis shows for the displacements a maximum deviation of 2.0 mm for values with a maximum of 18.7mm. Furthermore, the comparison of the cross-sectional forces and moments and the comparison of the contact forces show a sufficient agreement. All results of the benchmark analysis show values within the expected limits. I. e. the results gained by the benchmark analysis are within the margins (30%) defined for the design. In such a complex and multiple nonlinear structure, the benchmark analysis is the basis to get confident results since actual experimental values are not available.

REFERENCES 1. Jaksic N. et al., Definition der Randbedingungen bei einer FEStrukturanalyse durch Nutzung Symmetriebedingungen einer Stellarstoranordnung, IPP/Z3, february 1997 2. Bykov V. et al., Structural analysis of W7-X: Main results and critical issues, Fusion Engineering and Design 82 (2007), pp. 1538-1548. 3. Jaksic N. et al., Design analysis of the support structure stressed by large superconducting coils for a plasma fusion experiment, Computers & Structures 81 (2003) pp.697-714 4. Simon-Weidner J. et al., On mechanical stress in large helias coil systems and the influence of contact effects, proceedings of the 16th SOFT, London, (UK), 3-7 September 1990, pp. 1511-1514.

SIMULATION OF COMPLEX NON-LINEAR STRUCTURES IN LARGE SCALE ANALYSIS 5. Bathe K. J., Finite Element Procedures, Prentice Hall, Upper Saddle River, NJ 07458, 1996. 6. ADINA R&D Inc., A finite element computer program for Automatic Dynamic Incremental Nonlinear Analysis - System 8.5, 71 Elton Avenue, Watertown, MA 02472 USA, 2001. 7. ANSYS, Engineering Analysis System, Release 10.0A1, ANSYS Inc.

You might also like