From CAD Geometry to Defensible FE Model

What Is It?

The workflow from CAD geometry to a defensible finite element model is the complete process of turning the design geometry — the CAD model created by the design team — into an analysis model that is efficient, verified and technically defensible. The workflow covers every step: understanding the engineering question, extracting the relevant geometry, idealising the geometry (removing detail, abstracting dimensions), choosing the element type and mesh, defining the materials, applying the loads and constraints, verifying the model, running the analysis and documenting the result. Each step involves engineering judgement — what to include, what to omit, what to simplify — and each judgement should be recorded so that the model can be reviewed and defended. The workflow is not a recipe; it is a structured process that ensures the model is built for the right question, verified before the results are believed and documented so that the analysis is traceable.

Why It Matters

A defensible model is one that can withstand technical scrutiny — by a reviewer, by a certifying authority, by a peer, by a court. It is a model whose assumptions are explicit, whose idealisations are justified, whose verification is documented and whose results are traceable to the inputs. A model that is not defensible is a model whose results cannot be trusted under scrutiny — and in safety-critical work, scrutiny is inevitable. The workflow from CAD to defensible model ensures that every modelling decision is made deliberately, recorded and justified. It prevents the most common failure of FE analysis: a model that was built quickly, not verified, not documented and whose results are then used to make a critical decision that later turns out to be wrong. The workflow is the difference between an analysis that is trusted and one that is not.

A DEFENSIBLE MODEL IS ONE THAT CAN WITHSTAND TECHNICAL SCRUTINY. The assumptions are explicit, the idealisations are justified, the verification is documented and the results are traceable. The workflow from CAD to defensible model ensures that every modelling decision is made deliberately and recorded. In safety-critical work, a model that cannot be defended is a model whose results cannot be used.

Step 1 — Understand the Engineering Question

The first step is to understand the engineering question that the analysis must answer. Is the question a global stiffness question (does the structure deflect within limits)? A local stress question (what is the peak stress at the fitting)? A buckling question (what is the critical load)? A dynamic question (what is the natural frequency, the response to vibration)? A fatigue question (what is the stress at the critical detail for a fatigue life calculation)? The question determines the idealisation: a global stiffness model can be coarse and abstracted; a local stress model must be fine and detailed at the feature; a buckling model must capture the instability mode; a dynamic model must have the right mass and stiffness; a fatigue model must capture the stress concentration. The question also determines the required accuracy: a 10% error may be acceptable for a global deflection check but unacceptable for a fatigue life calculation. The engineering question should be stated explicitly before any modelling begins — and the model should be built to answer that question, not to be a general-purpose replica of the structure.

Step 2 — Extract and Idealise the Geometry

The second step is to extract the relevant geometry from the CAD and idealise it. The relevant geometry is the geometry that affects the engineering question — the load path, the stiffness, the mass, the local stress. The irrelevant geometry — cosmetic features, small chamfers, threads, detail that does not affect the question — is removed. The idealisation may also abstract the dimension: a slender beam becomes a 1D beam element; a thin wall becomes a 2D shell element; a complex mass becomes a 0D mass element. The idealisation should be driven by the structural behaviour, not by the CAD appearance. The idealised geometry should be a clean, analysis-suitable representation — not the CAD with a few features suppressed. Every idealisation decision (what was removed, what was abstracted, why) should be recorded for the documentation.

• Extract only the geometry that affects the engineering question
• Remove cosmetic features, small chamfers, threads, irrelevant detail
• Abstract dimensions: beam for slender, shell for thin, mass for equipment
• Create mid-surfaces for shell models; centre-lines for beam models
• Record every idealisation decision and the reason for it

Step 3 — Choose Element Type and Mesh

The third step is to choose the element type and the mesh density. The element type (beam, shell, solid; first-order, second-order; reduced integration, full integration) is chosen based on the structural behaviour and the stress requirement. The mesh density is chosen based on the stress gradient — fine where the stress varies rapidly (stress concentrations, fillets, holes), coarse where the stress is smooth (global regions). The mesh transition between fine and coarse regions should be gradual (no sudden size jumps). The element quality (aspect ratio, skew, Jacobian) should be checked after meshing. The mesh should be refined at the features that drive the result — a fatigue-critical fillet needs a fine mesh with second-order elements; a global deflection model can use a coarse mesh with first-order elements. The mesh density and the element type should be justified by a convergence check — refine the mesh and confirm the result does not change.

Step 4 — Materials, Loads and Constraints

The fourth step is to define the materials, the loads and the constraints. The materials are assigned from the material specification — the elastic modulus, the Poisson's ratio, the density, the yield stress, the allowables. The material properties should be in a consistent unit system and should be checked against the specification. The loads are applied based on the load case definition — the external forces, pressures, temperatures, accelerations. The loads should be applied at the correct location with the correct distribution (distributed, not point, where local stress matters). The constraints are applied based on the support definition — the fixed, pinned, sliding or symmetry conditions. The constraints must prevent rigid-body motion and must represent the physical supports. The loads and constraints should be verified by the reaction balance and the deformation shape. Every material, load and constraint should be traceable to its source (material spec, load case document, support drawing).

Step 5 — Verify the Model

The fifth step is to verify the model before running the production analysis. The verification follows the checklist from the model verification article: check the mass, the CG, the equilibrium, the connectivity, the constraints, the reactions, the deformation shape, the displacement magnitude, the stress reasonableness, the modal behaviour and the element quality. Each check should be performed and the result recorded. If any check fails, the model is corrected and re-verified. The verification is the gate — no production analysis is run until the model passes all the verification checks. The verification is not optional and not a formality; it is the process that catches the errors that would otherwise produce wrong results. A model that has not been verified is a model whose results cannot be trusted.

VERIFICATION IS THE GATE. No production analysis is run until the model passes all the verification checks — mass, CG, equilibrium, connectivity, constraints, reactions, deformation, displacement, stress, modal and element quality. Each check is performed and recorded. If any check fails, the model is corrected and re-verified. Verification is not optional.

Step 6 — Run the Analysis and Check the Results

The sixth step is to run the production analysis and check the results. The results should be checked for the same qualities as the verification: the deformation shape should be physical, the reactions should balance, the stress should be reasonable. The results should also be checked for convergence — refine the mesh and confirm the peak stress does not change. The results should be examined in the context of the engineering question: does the deflection meet the limit? Is the peak stress below the allowable? Is the buckling load above the applied load? Is the natural frequency away from the excitation? The results should be compared with hand calculations, with previous analyses and with test data where available. A result that contradicts a hand calculation or a physical expectation should be investigated — not dismissed and not accepted without understanding.

Step 7 — Document and Make Defensible

The seventh step is to document the model and the analysis so that the result is defensible. The documentation should include: the engineering question and the required accuracy; the geometry source and the idealisation decisions; the element type and mesh (with the convergence check); the materials (with the source and the unit system); the loads and constraints (with the source and the verification); the verification results (all checks, pass or fail); the analysis results (with the comparison to allowables and to hand calculations); and the conclusions. The documentation should be sufficient for a reviewer to understand the model, to repeat the verification and to trace every result to its input. A well-documented analysis is one that can be reviewed, repeated and defended — even years later, by someone who was not involved in the original work.

Workflow: CAD geometry to defensible FE model

  ┌──────────────────────────────────────────────────────┐
  │  1. Engineering question + required accuracy          │
  │     ↓                                                 │
  │  2. Extract + idealise geometry                       │
  │     (remove detail, abstract dimension)               │
  │     ↓                                                 │
  │  3. Element type + mesh (convergence check)           │
  │     ↓                                                 │
  │  4. Materials + loads + constraints (traceable)       │
  │     ↓                                                 │
  │  5. VERIFY (mass, CG, reactions, deformation, modal)  │
  │     ↓  (gate — no production run until verified)      │
  │  6. Run analysis + check results                      │
  │     (physical, balanced, converged, compared)         │
  │     ↓                                                 │
  │  7. Document (assumptions, verification, results)     │
  │     → defensible, traceable, reviewable               │
  └──────────────────────────────────────────────────────┘

  Each step involves engineering judgement — recorded, not implicit.

Traceability

Traceability is the ability to follow every result back to its source — the CAD version, the material specification, the load case document, the support drawing, the verification check. A traceable model is one where a reviewer can pick any number in the results (a peak stress, a displacement, a natural frequency) and trace it back through the model to the input that produced it. Traceability is achieved by recording the source of every input, the version of every file and the result of every check. It is the foundation of defensibility — a model without traceability is a model where the results cannot be audited. In safety-critical work, traceability is not optional; it is a requirement. The documentation should include a traceability matrix that links each result to its inputs and its verification.

IDEALISATION CONSIDERATION: The workflow is not linear — it is iterative. The idealisation may change when the verification reveals an issue; the mesh may change when the convergence check shows non-convergence; the loads may change when the reaction balance fails. The workflow is a loop of build, verify, correct, re-verify — until the model is correct and the results are trusted. Do not treat the workflow as a one-pass recipe; treat it as an iterative process of refinement.

Key Takeaways

• The workflow covers: question, geometry, idealisation, mesh, materials, loads, constraints, verification, analysis, documentation
• The engineering question drives the idealisation — state it explicitly before modelling
• Every idealisation decision should be recorded — what was removed, what was abstracted, why
• Verification is the gate — no production analysis until the model passes all checks
• Documentation makes the model defensible — assumptions, verification, results, traceability
• The workflow is iterative — build, verify, correct, re-verify until the model is correct