Realistic Large-scale Industrial Analyses

 The Nafems Education and Training Working Group (ETWG)  wish to establish a project that will:

1. define a number of realistic industrial analysis problems and
2. show solutions to each problem using different modelling techniques (and programs perhaps).

The primary objective of this project is to illustrate the main issues related to achieving an appropriate and known level of accuracy for large, realistic structural analyses.

Work Required

Stage 1 – Problem Definition

• Define a set of static or quasi-static structural problems (dimensions, materials, loads, boundary conditions) and the required results.
• Ideally there should be 3-5 separate models, with a range of behavioural types needing a range of element types for the solution(s). At least one should have a mix of “shell-like” and “chunky” features.
• Highlight the potential pitfalls, issues to be aware of, F.E. modelling options to consider.
• Most likely the geometry will need to be defined in computer files and these must be readily available.
• Issue a booklet describing all of the above and promote awareness of the tasks.

Stage 2 –Analysis

• Run a number of analyses for each model, possibly using using different behavioural types (e.g. shells or 3D continua), different idealisations (element types and mesh), different analysis programs and different analysis assumptions (like assumed linearity compared with geometric non-linearity). Alternative computational techniques to the FE method may be used. These solutions should be such that they highlight the importance of good practice for element selection and meshing, drawing attention to common problems and pitfalls (e.g. the sensitivity of tets to distortion) and illustrate the issue of balancing accuracy against computational resource, with respect to structural representation and element selection, as well as modelling techniques such as mesh adaptivity, submodelling, and so on.
• It is anticipated that the models should end up with no more than 3 million  DOF, to ensure that they can be run on standard computers.
• Outline conclusions on accuracy, complexity, computational resource.
• Issue a bookdescribing all of the above.
• Compile a presentation summarising the book.

Intended Readership

• FE analysts who are getting past the beginner stage.
• Experienced analysts who are broadening the width of analysis types with which they are familiar.
• “Designer-analysts” using FE as an adjunct to the CAD program.

Tender

The project involves industrial application of analysis technology and so should ideally be performed by an industrial organisation or consortium that has experience of real-world structural analysis and is headed up by a well qualified individual.

 It may be that more than one tender will be accepted.

 The tender document should include:

•  an outline of the proposed range of structural models,
• how the models will be defined (e.g. STEP 203 geometry, plus drawings and text),
• the work plan and costs for managing the two parts of the project and producing the deliverables. The total cost should not exceed £4,000.
• the potential partners who will contribute to the Part 1 exercise and run some analyses,
• proposals for dissemination outside the Nafems “analyst” community, reaching the “designer-analyst” community.
• the head candidate’s CV, highlighting experience in carrying out industrial structural analysis on real-world components or structures. It may be useful to provide CVs of other key people.

Format

Instructions on the format(s) to be used for the deliverables will be provided by Nafems.

Important Notes

• It is not the objective to highlight deficiencies in any particular software or in any group of individuals.  A certain degree of anonymity may therefore be required when reporting submitted results.
• The objectives of the exercise are to create awareness of issues related to the accuracy of structural analysis, to promote best practice in terms of model construction, and to allow NAFEMS members to participate in the exercise and benefit from the collective experience of analysts, vendors and academics.
• It is also hoped to attract the attention of “designer-analysts” currently outside Nafems.
• The project may well promote discussion amongst experts, vendors, academics and analysts, but it should be stated from the outset that there will be no ‘right’ answer.
•  It may well be that Nafems will want to run a seminar as one way of disseminating results and engendering further discussion and work.

A Very Simple Illustrative Example

Stage 1

Here is a slender structure.

The loading is a pressure of magnitude X applied here.

The boundary conditions are encastre here.

The material is mild steel, modelled as purely elastic.

Devise suitable F.E. models using shell and 3D continuum elements to predict stress at this location (XYZ) and displacement at this location (XYZ).

Show how the results converge with mesh refinement or element type.

Stage 2

The ‘reference’ solutions for stress and displacements are X and Y (obtained from a variety of codes).

If you model this structure with first-order reduced integration shells you get an answer which is X% in error. It is best to use second order elements.

If you model with elements larger than X you get Y% error. The mesh density needs to be around X in order to capture the stress gradients adequately.

If you model with 3D elements the displacements are about the same as those for shells, but the compute cost is much higher.

Comments about how to model more complex boundary conditions, load applications, material behaviour, etc. Also, discussion of other modelling methods (transitions between element types/topologies)

Next Steps & Closing Date

In the first instance, tender documents should be submitted to the working group chairman at etwg@nafems.org

Closing date for submissions is 15 Dec 2008.