KnowMesh
← Back to search

Anna Adamska

Senior Materials Engineer · Materials & Process Engineering

Warsaw, Poland

Senior Materials Engineer with a decade of experience in aerospace material characterization, allowables database development, and supplier qualification. Leads material selection and qualification activities for new propulsion programs — covering nickel superalloys, titanium alloys, and advanced coatings. Deep knowledge of statistical methods for design allowables and MMPDS/CMH-17 data reduction procedures.

Expertise

  • nickel superalloy and titanium alloy characterization
  • material allowables database development
  • supplier material qualification
  • design allowables for hot section components
  • material specification authoring (AMS, MMPDS)

Technologies

Thermo-Calc Python Minitab MMPDS AS9100 Excel (VBA) Git

Work History

2025-01

AI-assisted material screening tool — collaborated with AI team to build a property prediction model for new alloy compositions. Provided 800 historical alloy characterization data points as training data and validated model predictions.

Challenge: Historical alloy data had significant batch-to-batch variability that the AI model treated as feature variability rather than measurement noise. Clustering by alloy heat number and averaging within heats before training reduced prediction error by 35%.

Learned: Material property ML models must account for the hierarchical data structure — batch-level vs. specimen-level variability. Feeding raw specimen data without heat-level grouping mixes process variation with material property variation, degrading model accuracy.

Python pandas

2024-04

Thermodynamic equilibrium calculations for new γ' strengthened superalloy — used Thermo-Calc with TCNI9 database to predict phase stability, γ' solvus temperature, and TCP phase formation risk as a function of composition.

Challenge: TCNI9 database predictions for Re content above 5% showed significant uncertainty in TCP phase formation temperature. Required experimental validation — commissioned 6 alloy buttons at compositions spanning the design space for DSC measurement of actual solvus temperatures.

Learned: Thermo-Calc predictions for high-Re superalloy compositions must be validated experimentally. The database accuracy for complex multi-element superalloys decreases significantly above 3-4 alloying element interactions — computational predictions are a useful starting point, not a design basis.

Thermo-Calc Python MATLAB

2023-09

New forging supplier qualification — defined the qualification test plan, reviewed supplier qualification specimens, and approved the material certification against AMS 5596 for IN718 HPT disk forgings.

Challenge: Supplier's initial qualification specimens had grain size non-conformances in the bore region — ASTM 5 required but supplier achieved ASTM 4 in 30% of specimens. Negotiated a corrective action plan with additional forging process optimization and re-qualification testing before approval.

Learned: Supplier material qualification must include microstructural acceptance criteria alongside mechanical property tests. Mechanical properties can meet specification even with suboptimal microstructure — but the microstructure controls scatter and long-term durability in ways that short-term testing cannot detect.

Python Minitab AS9100

2023-02

Design allowables database development for IN718 forged HPT disk — coordinated test campaign with mechanical test lab (120 tensile, LCF, creep specimens). Performed statistical data reduction per MMPDS-10 for A- and B-basis allowables.

Challenge: LCF data at 650°C showed bimodal scatter — two distinct failure populations with different crack initiation mechanisms (surface vs. subsurface). MMPDS standard assumes a single Weibull distribution. Required a mixture model approach and DER justification for the non-standard data reduction method.

Learned: Bimodal fatigue scatter is a common challenge for nickel superalloys with surface and subsurface crack initiation competition. Document the failure mode population separately — a single-distribution allowable calculated over mixed populations will be unconservative for one failure mode.

Minitab Python MMPDS