Aleksandra Malinowska

2025-01

MAC matrix computation and mode tracking across design iterations of a bracket — automated Python script that reads Nastran f06 files, extracts eigenvectors, and computes MAC for up to 30 modes.

Challenge: Eigenvector sign is arbitrary in Nastran — MAC computation needs to handle sign ambiguity. Also, mode ordering can change between design iterations (mode veering), so simple index-based pairing fails.

Learned: pyNastran's OP2 reader is the right tool for automated Nastran post-processing. Parsing f06 text files is fragile and error-prone. MAC-based mode pairing handles sign ambiguity and mode veering correctly.

2024-07

Modal effective mass table generation for equipment mounting bracket — calculated modal effective mass and cumulative effective mass in all 6 directions to identify significant modes for each direction.

Challenge: Cumulative effective mass in the lateral direction reached only 78% (target >90%) with 50 modes — indicating significant contribution from higher modes. Extended modal extraction to 100 modes and the cumulative reached 94%.

Learned: Modal effective mass is the key metric for deciding how many modes to include in a dynamic analysis. Rules of thumb (e.g., '3x highest frequency of interest') are unreliable for complex structures — compute effective mass explicitly.

2024-02

Craig-Bampton Component Mode Synthesis (CMS) — reduced HPT module to a superelement for use in full engine system dynamics model. Verified superelement accuracy by comparing full and reduced model natural frequencies.

Challenge: Superelement boundary conditions definition — automatic ASET generation by Nastran selected non-physical boundary nodes. Manual ASET definition based on physical interface geometry gave more accurate reduced model.

Learned: Nastran's automatic ASET (attachment set) generation for CMS is not always optimal. For complex component interfaces, manual definition of physically meaningful interface nodes gives better reduced model accuracy.

2023-11

Shock response spectrum (SRS) analysis for pyrotechnic separation event — transient modal analysis SOL 112, extracted SRS at critical mounting points, compared with equipment qualification envelope.

Challenge: SRS calculation from transient results requires careful selection of peak-picking method (maximax SRS). Implemented maximax SRS post-processing in MATLAB since Nastran does not natively output SRS.

Learned: SRS is a derived quantity — the analysis gives time history, and SRS computation is post-processing. Always validate the SRS calculation implementation against a known analytical solution before applying to real problems.

2023-06

SOL 111 modal frequency response for avionics mounting bracket — transfer functions from engine mount excitation to avionics mounting points, comparison with vibration qualification test envelope.

Challenge: Direct vs modal frequency response (SOL 108 vs SOL 111) gave 15% different results at the first resonance. Root cause: modal truncation — higher modes not included in SOL 111 model were contributing to response at the frequency of interest.

Learned: Modal truncation error in SOL 111 is problem-specific. For response dominated by high-frequency modes, including residual inertia correction (PARAM,RESVEC,YES in Nastran) or switching to SOL 108 direct method is necessary.

2023-01

SOL 103 normal modes analysis for engine nacelle — extracted 50 mode shapes and natural frequencies, compared with ground vibration test (GVT) data using MAC matrix and frequency deviation criteria.

Challenge: MAC values below 0.7 for three modes in the 80-120 Hz range. Investigation showed that the bonded aluminium honeycomb core was modeled with overly stiff adhesive layer properties. Calibrating adhesive shear stiffness from coupon test data improved MAC to > 0.85 across all modes.

Learned: Sandwich panel dynamic behavior is dominated by the core shear stiffness and face sheet bending stiffness. Getting these right requires coupon-level material characterization, not just handbook values.