Karolina Dabrowska

2025-01

Developed Fortran UMAT for crystal plasticity model of single crystal nickel superalloy — 12 slip system Schmid-based plasticity with kinematic and isotropic hardening. Verified against tension test data along [001], [011], and [111] orientations.

Challenge: UMAT debugging is notoriously difficult — the subroutine is called thousands of times per increment and standard debugging tools don't work well in this context. Built a standalone Fortran test driver to verify the constitutive model response before integrating with Abaqus.

Learned: Always develop and validate UMAT logic in a standalone driver before integrating with Abaqus. The standalone driver allows unit testing of the constitutive update algorithm with known strain inputs and expected stress outputs.

2024-09

High-temperature creep analysis of HPT blade at takeoff and cruise conditions — Norton power law creep model, multi-step analysis with hold times for takeoff/cruise/descent cycle.

Challenge: Creep analysis with very different time scales (0.5h takeoff vs 5h cruise hold) required careful time stepping — small steps during fast initial creep, larger steps during steady-state creep. Abaqus's automatic incrementation worked well with CETOL tolerance set to 1e-4.

Learned: Creep analysis is one of the cases where Abaqus automatic incrementation is genuinely useful — the solver adapts step size based on creep strain rate. Manual time stepping is rarely better. Key is setting CETOL correctly for the expected creep rate.

2024-04

Fretting fatigue analysis at HPT blade dovetail attachment — cyclic contact between blade root and disk slot, microslip regime, contact stress cycles as input to fretting fatigue criterion (SWT parameter).

Challenge: The dovetail contact slip regime was sensitive to friction coefficient — 0.25 vs 0.30 changed the predicted microslip amplitude by 40%. Measured friction coefficient from fretting rig tests was essential input.

Learned: Fretting fatigue predictions are only as good as the friction coefficient data. Handbook values are insufficient for tight contact geometries — test-derived friction data is necessary for reliable life prediction.

2023-11

Threaded connection analysis — bolt shank, thread engagement region, and nut modeled with helical thread geometry and friction contact. Bolt preload applied via temperature analogy in Abaqus.

Challenge: Helix angle causes asymmetric load distribution in thread engagement — leading thread bears 30-40% of total load. Meshing the helical thread geometry in HyperMesh was time-consuming — helical extrusion required manual layer-by-layer meshing.

Learned: Detailed thread geometry models are expensive to build and run. For most bolt joint analyses, simplified models (no thread detail, BEAM or connector elements) are sufficient. Full thread models are only justified when thread failure or wear is the primary concern.

2023-06

Rubber O-ring compression analysis using hyperelastic material (Mooney-Rivlin model, constants calibrated from uniaxial tension and equibiaxial tension test data). Predicted contact width and sealing pressure vs compression ratio.

Challenge: Mooney-Rivlin model became unstable at large compressive strains (>50%) — negative Drucker stability determinant. Switched to Yeoh model which is stable across a wider strain range and fitted the test data comparably well.

Learned: Hyperelastic model selection matters enormously. Always check Drucker stability across the full expected strain range before using a hyperelastic model in a contact problem. An unstable model will produce non-physical results without obvious error messages.

2023-01

Press fit analysis with radial interference — turbine disk bore and shaft. Elastic-plastic analysis with nonlinear geometry to capture redistribution of contact stress under operating temperature.

Challenge: Contact pressure distribution was non-uniform due to end effects — peak contact stress at the edge was 2.3x the nominal. Needed to capture this accurately for fretting fatigue assessment. Required mesh refinement at the interference boundary.

Learned: Interference fit contact stress edge effects are always present and often govern fatigue life. Classical Lame solution underestimates edge peaks by 2-3x. FEM is necessary for accurate edge stress prediction.