Grzegorz Lewicki

2024-11

Pre-test prediction for static ultimate load test of wing spar — predicted failure load, failure location, and deformation at limit load for comparison with test measurements.

Challenge: Pre-test prediction had to be submitted 3 weeks before the test without knowledge of test fixture stiffness. Ran sensitivity study with ±20% fixture stiffness — maximum displacement varied by 12%, but critical margin was insensitive. Documented this quantitatively in the prediction report.

Learned: Pre-test predictions must include quantified uncertainty bounds for all assumed boundary conditions. A single-point prediction without sensitivity analysis is insufficient for meaningful test-analysis correlation.

2024-05

SOL 200 size and shape optimization of stiffened panel — minimize weight subject to stress, buckling, and deflection constraints across 8 load cases. 42 design variables (sheet thickness, stiffener dimensions).

Challenge: SOL 200 convergence was poor — optimizer was cycling between two design points without progressing. Root cause: gradient inversion near a constraint boundary. Adjusted design variable bounds and added a soft constraint on buckling load to smooth the design space.

Learned: SOL 200 optimization is sensitive to the constraint formulation and design variable scaling. Constraint functions with very different magnitudes (MPa stresses vs dimensionless buckling factors) cause gradient scaling issues. Normalize all constraints to a common scale.

2023-11

Inertia relief analysis (SOL 101 with PARAM,INREL,-2) for free-free engine component under combined thrust, gyroscopic, and pressure loading — acceleration field balanced by distributed inertia forces.

Challenge: INREL analysis requires exactly 6 degrees of freedom unrestrained — more or fewer causes a singular stiffness matrix. The component had symmetry constraints applied for visualization that accidentally over-constrained it.

Learned: Inertia relief is sensitive to the definition of the free-free boundary condition. A common mistake is applying symmetry constraints that are physically wrong for the free-free case. Always verify with a rigid body motion check before running.

2023-07

SOL 105 linear bifurcation buckling analysis of composite compression panel — extracted first 10 eigenvalues and mode shapes, identified critical buckling mode (local skin buckling at 1.65 design limit load).

Challenge: Several eigenvalues were very close in value (within 2%) — different from expected well-separated buckling modes. Root cause: poorly constrained out-of-plane rigid body motion in the model caused spurious near-zero eigenvalues. Added SUPORT to properly constrain RBM.

Learned: Spurious low eigenvalues in SOL 105 almost always indicate either unconstrained rigid body modes or very flexible elements (weak springs). Always check the first few eigenvectors visually before reporting buckling factors.

2023-02

SOL 101 linear static analysis of wing attachment fitting under ultimate load combinations — 12 load cases from loads envelope, margin of safety extraction, and fastener load distribution.

Challenge: Extracting fastener loads from CBUSH elements for all 12 load cases and combining them into a single MOS table required a Python post-processing script. Nastran's native f06 output was impractical for this scale.

Learned: For large Nastran models with many fasteners and load cases, automated result extraction via the Nastran OP2 binary interface (using pyNastran) is orders of magnitude faster than parsing f06 text output.