FEM-Study ILD-AHCAL. Seismic simulation. K. Gadow, M. Lemke, F. Sefkow Santander,

Size: px

Start display at page:

Download "FEM-Study ILD-AHCAL. Seismic simulation. K. Gadow, M. Lemke, F. Sefkow Santander,"

Laurence Warren
5 years ago
Views:

1 FEM-Study ILD-AHCAL Seismic simulation K. Gadow, M. Lemke, F. Sefkow Santander,

2 Overview > Reminder of earlier studies > Sub-structuring method > First results > Outlook Martin Lemke FEM-Study ILD-AHCAL Seite 2

crane sub module connection by plates from the front

3 AHCAL submodule case 2 submodule installation installation of submodules in front of the cryostat by crane sub module connection by plates from the front and back side Martin Lemke FEM-Study ILD-AHCAL Seite 3

Earlier studies > Static calculations of absorber structure deformation and stress

preprocessed for dynamical analysis > First step: modal analysis: determine eigen

very challenging Full model impractical, slow convergence For the study of the

4 Earlier studies > Static calculations of absorber structure deformation and stress Including configurations during assembly Design validated > Data from 2011 earthquake preprocessed for dynamical analysis > First step: modal analysis: determine eigen modes and frequencies of system > Combination of large dimensions and small details very challenging Full model impractical, slow convergence For the study of the complete configuration a different approach is needed Martin Lemke FEM-Study ILD-AHCAL Seite 4

5 AHCAL Mechanics absorber structure validation (static) ~360 N/mm² maximum tension 50 times exaltation graphic Martin Lemke FEM-Study ILD-AHCAL Seite 5

Data preparation for Response Spectrum Analysis >

prepare data > Tool to generate power spectrum as input

screenshot and Open Source Tool Get Data Graph Digitizer

6 Data preparation for Response Spectrum Analysis > NIED-Software Tool Strong Motion Data Analysis used to prepare data > Tool to generate power spectrum as input for response spectrum Analysis > Data extraction via screenshot and Open Source Tool Get Data Graph Digitizer and export to Excel Martin Lemke FEM-Study ILD-AHCAL Seite 6

7 Modal analysis: Eigen modes > Swinging barrel: 3Hz > Swinging module: 8Hz > Swinging plate: 6Hz > Higher modes: 15 Hz > Several plates: 45 Hz Martin Lemke FEM-Study ILD-AHCAL Seite 7

8 Appropiate calculation method for complex structures > First tests with a substitution method for the complex AHCAL-Segments with effective material parameters (effective elastic modules and sheer modules, as well as effective Poisson's ratio) in combination with a homogeneous body as a replacement for the detailed AHCAL-Segment => unfortunately no real breakthrough > Another way to calculate such a complex model like the AHCAL- Structure had to be found > The chosen calculation method is the Substructuring Method and hereby especially the Component Mode Synthesis (CMS) as a form of substructure coupling analysis in ANSYS used in structural dynamics > Built up test cases in ANSYS to develop the APDL-Command-Snippets (handling geometry/model, combine result-files) Martin Lemke FEM-Study ILD-AHCAL Seite 8

Substructuring A short explanation > Substructuring is a procedure that condenses a group of finite elements into one element represented as a matrix.

9 Substructuring A short explanation > Substructuring is a procedure that condenses a group of finite elements into one element represented as a matrix. The single-matrix element is called a superelement. (Source: ANSYS Help V17.0) > Three steps to use this method: Generation Pass to calculate the superelement matrices (mass, stiffness, damping) and to define the Master DOFs Use Pass to integrate the superelments into the Finite Element Model, the Master DOFs are used to define the boundary conditions (supports, loads, contacts) Expansion Pass to assign the results of the superlements to the original geometries Figure-Source: CADFEM GmbH/Germany Martin Lemke FEM-Study ILD-AHCAL Seite 9

10 Scheme for dynamic analysis-run with CMS-Method Build up CMS-Model with Generation and Use Pass > Generation Pass only calculated for the first run, with same model, the generated Superelements (and the corresponding files) can reused in various analysis again Geometry Preparation Static Analysis Prestress Modal Analysis MSUP Response Spectrum Analysis View Results Expand Superlement- Results onto original Structures Expand Superlement- Results onto original Structures Expand Superlement- Results onto original Structures Martin Lemke FEM-Study ILD-AHCAL Seite 10

11 Substructuring Pros and Cons > Pro: Complex models can be calculated, n-elements describing the complex FE-model can be simplified with the CMS-method to one! single element For same geometry/mesh, the created files for super-elements can be saved/reused (huge time saving potential) Very good performance (with reusing super-element-files) in dynamic analysis Less relevant eigen modes can be sorted out by filtering the calculated modes with APDL-Commands > Contra: Long time to simplify geometry / to prepare the Finite Element Model Long preparation/calculation time in Generation Pass and Expansion Pass (these are additional calculations to the real model/analysis) => BUT: very effective in complex models with several analysis steps APDL (ANSYS Parametric Design Language) and FE-models around APDLcommands (sometimes) sensitive to handle ;-) Martin Lemke FEM-Study ILD-AHCAL Seite 11

Towards the whole barrel-setup (long(er) term) > To get real amplitudes, the whole support-structure for the AHCAL has to be calculated > There the CMS-Method is the key => a possible path: In the

12 Towards the whole barrel-setup (long(er) term) > To get real amplitudes, the whole support-structure for the AHCAL has to be calculated > There the CMS-Method is the key => a possible path: In the first step, the barrel yoke with feet and cryostat, and mass points for AHCAL+ECAL+ will be calculated Second: the system response at the AHCAL guide rails will be analysed and saved Third: with these detailed excitations at AHCAL guide rails a second analysis turn will be done for the AHCAL (again with CMS-Method) > For the time being, concentrate on AHCAL and excite rails directly Martin Lemke FEM-Study ILD-AHCAL Seite 12

to detailed AHCAL geometry > Geometry and mesh for 3D-Full-Model and

13 Tests to develop the substructuring process for AHCAL > Simple model, only 8 segments connected by front plates => geometrically identical to detailed AHCAL geometry > Geometry and mesh for 3D-Full-Model and CMS-Model identical Martin Lemke FEM-Study ILD-AHCAL Seite 13

14 Tests to develop the substructuring process for AHCAL > Boundary conditions comparable to AHCAL-boundary conditions (prism-supported side and loose side, free in x-direction) identically as well Martin Lemke FEM-Study ILD-AHCAL Seite 14

15 Tests to develop the substructuring process for AHCAL > Modal results of 3D full model Mode f [in Hz] 1 119, , , , , ,63 Martin Lemke FEM-Study ILD-AHCAL Seite 15

16 Tests to develop the substructuring process for AHCAL > Modal Results of CMS-Model Mode f [in Hz] 1 118, , , , , ,89 Martin Lemke FEM-Study ILD-AHCAL Seite 16

performed to check the calculation accuracy of FE

17 Tests to develop the substructuring process for AHCAL > Single static-mechanical and modal analyses performed to check the calculation accuracy of FE models using the CMSmethod > Fixed support and two forces > 6,4 Mio. elements in both analyses (3D-/CMS-FE model) Martin Lemke FEM-Study ILD-AHCAL Seite 17

18 Tests to develop the sub-structuring process for AHCAL > Accurate results with CMS method 3D-Full-Model CMS-Model Martin Lemke FEM-Study ILD-AHCAL Seite 18

19 Tests to develop the sub-structuring process for AHCAL > Accurate results with CMS method 3D-Full-Model CMS-Model > Should be verified by measurement Martin Lemke FEM-Study ILD-AHCAL Seite 19

20 General AHCAL-Model > Results CMS-Model: Shell-Model (prestressed) Shell-Model (free) CMS-Model (free) Nr. f [in Hz] Nr. f [in Hz] Nr. f [in Hz] 1 2,97 1 2,83 1 3,13 2 5,27 2 5,18 2 5,73 3 6,11 3 7,07 3 8,06 4 7,65 4 7,46 4 8,41 5 9,16 5 9, ,84 6 9, , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , ,04 Martin Lemke FEM-Study ILD-AHCAL Seite 20

21 General AHCAL-Model > Results 3D-Model: Shell-Model (prestressed) Shell-Model (free) CMS-Model (free) Nr. f [in Hz] Nr. f [in Hz] Nr. f [in Hz] 1 2,97 1 2,83 1 3,13 2 5,27 2 5,18 2 5,73 3 6,11 3 7,07 3 8,06 4 7,65 4 7,46 4 8,41 5 9,16 5 9, ,84 6 9, , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , ,04 Martin Lemke FEM-Study ILD-AHCAL Seite 21

22 Nice option: The Modal Participation Factor > Modal Participation Factor to sort out irrelevant eigenmodes; especially the AHCAL-Structure produces a lot of eigenmodes caused by large amount of single plates in each of 16 segments > This method saves disc space and reduces calculation time > Only interesting modes used for following dynamic analysis with MSUP- Method Martin Lemke FEM-Study ILD-AHCAL Seite 22

23 To Do: General tasks (short term schedule) > Build up the pre-stressed modal analysis with CMS-Testmodel (actually working) > Response spectrum analyses with earthquake data generalised with ISO 3010 and Eurocode 8 > Handover the modal-results to a response spectrum analysis with Earthquake data from Japan/Ichinoseki (Source:NIED K-NET IWT /03/11) has to be implemented in the Testmodel, max. amplitudes: East-West 852,134 gal (alternatively 8,52 m/sec² => 0,36 g) North-South: 997,780 gal (alternatively 9,98 m/sec² => 1,02 g) Vertical: 352,666 gal (alternatively 3,53 m/sec² => 0,36 g) > Mandatory: regular adjustments in models (Commands, geometry, ) Martin Lemke FEM-Study ILD-AHCAL Seite 23

24 Response Spectrum Analysis > In DIN 3010, the Response Spectrum Analysis ist recommended for earthquake analysis on structures > In ANSYS the Response Spectrum Analysis uses the excitation power spectrum to calculate the maximum answer of the structure (max. amplitude) for a specific frequency > For Response spectrum analysis the SRSS-Method is used (SRSS => Square root of sum of squares) > Mass- and stiffness damping (both 0,01) are used to calculate real/finite amplitudes for the structure > The frequency at the max. Amplitude in Response Spectrum Analysis con be used to set up a Transient Analysis to investigate timedependent behavior of the structure Martin Lemke FEM-Study ILD-AHCAL Seite 24

25 Damping in Model > Damping in FE-Method is not trival > Mass-/Alpha- und Stiffness-/Beta-Damping are aproximations to to introduce the damping into the FE-Method (one possibility) and is called as Rayleigh Damping > The Damping is difened as following: [C] = α[m] + β[k] with: [M] => Massmatrix [K] => Stiffnessmatrix > The graphical summary (Source: ANSYS Help V17.0): > Damping Ratio defined as a sum of Alpha and Beta Damping over the frequency Martin Lemke FEM-Study ILD-AHCAL Seite 25

26 Pre-stressed Testmodel with CMS-Method > Setup a pre-stressed, damped Modal Analysis > Test setup (simple ring of eight segments) used to develop ANSYS program scripts for data transfer from Static Structural Analysis to Modal Analysis afterwards to Response Spectrum Analysis > In the Static Structural Run the load dependent structure deformations are calculated > The Modal Analysis in ANSYS uses the linear Perturbation Method to calculate the initial deformations of the structure > Successfully tested and adaption to the AHCAL-FE-Model > To reduce the complexity of the AHCAL-FE-Model, only the AHCAL- Structure is used in the analysis without barrel, cryostat, > Actually the pre-stressed AHCAL-Model is in calculation since one day Martin Lemke FEM-Study ILD-AHCAL Seite 26

27 Conclusions for prestressed AHCAL-Model with CMS- Method > Difficult to debug the complete AHCAL-FE-Model-Setup > Actually the scripts and the model works well > The actual model size is about 1.5 Mio Nodes > 95 GB RAM are used by the Block anzos Modal Solver in ANSYS > Temporary Solver files have a size around 400 GB!!! > The solution is very slow caused by a lot I/O (data file handling) within the solve process (=> different FEM-Server-Setup needed to optimize the solution-process?!) Martin Lemke FEM-Study ILD-AHCAL Seite 27

28 Alternative Setup for AHCAL-Structure Geometry > Ring-structures between the two AHCAL-Elements are introduced to stiffen the AHCAL-Structure built up with 16 segments > Each ring has a thickness about 15mm > Frontview: Martin Lemke FEM-Study ILD-AHCAL Seite 28

29 Alternative Setup for AHCAL-Structure Geometry > Backview: Martin Lemke FEM-Study ILD-AHCAL Seite 29

AHCAL-Geometry > The complete Geometry is built as

30 Alternative Setup for AHCAL-Structure Mesh/Model > The ring structrue is contacted to the original AHCAL-Geometry > The complete Geometry is built as shells Martin Lemke FEM-Study ILD-AHCAL Seite 30

31 Alternative Setup for AHCAL-Structure Results > The ring structure stiffened the AHCAL Geometry > Deformations about 4,5 mm reduced to 50% compared to original AHCAL Geometry Concept > A slight assymtry is present > Development of mountable ring structure for AHCAL-Front Martin Lemke FEM-Study ILD-AHCAL Seite 31

32 Alternative Setup for AHCAL-Structure Results > Graphics 50 times inflated Martin Lemke FEM-Study ILD-AHCAL Seite 32

33 Summary, outlook and remarks > Seismic study is progressing > Successful change to sub-structure method which will ultimately allow integration into cryostat and yoke whilst retaining all details > Getting ready to simulate with real earthquake data > However, do far each new step brought new challenges > Will also address rotated case (by 22 o as preferred by ECAL) and ECAL support details > Results are conservative with respect to Thicker absorber (fewer layers for cost resuction) Smaller inner radius Adaption to SiD Martin Lemke FEM-Study ILD-AHCAL Seite 33

34 Glossar > APDL => ANSYS Parametric Design Language (ANSYS scripting/ programming language to expand the capabilities in ANSYS Workbench) > Master DOF => a previously defined set of nodes, which are excluded from substitution with the superelement and to define boundary conditions and contacts to other bodies/parts/superelements > CMS => Componenent Mode Synthesis > MSUP => Modal Superposition Method > Modal Participation Factor => Each mode of a structure has an effective mass representing the share in the relevance for structural dynamic investigations, for this reason a modal participation factor less then a defined value will have no relevant impact on following dynamic analsis reusing the modal results for the MSUP-Method Martin Lemke FEM-Study ILD-AHCAL Seite 34

Chapter 5. Vibration Analysis. Workbench - Mechanical Introduction ANSYS, Inc. Proprietary 2009 ANSYS, Inc. All rights reserved.

Chapter 5. Vibration Analysis. Workbench - Mechanical Introduction ANSYS, Inc. Proprietary 2009 ANSYS, Inc. All rights reserved. Workbench - Mechanical Introduction 12.0 Chapter 5 Vibration Analysis 5-1 Chapter Overview In this chapter, performing free vibration analyses in Simulation will be covered. In Simulation, performing a