KnowMesh
← Back to search

Krzysztof Adamczyk

Senior Propulsion Engineer · Propulsion Engineering

Warsaw, Poland

Senior Propulsion Engineer with a decade of experience in gas turbine thermodynamic cycle analysis and engine performance modeling. Leads performance model development for new turbofan variants and deterioration assessments for in-service fleets. Primary toolset is NPSS and GSP, with Python automation for parametric studies and large-scale deck generation.

Expertise

  • gas turbine thermodynamic cycle analysis
  • engine performance modeling
  • SFC optimization
  • turbomachinery system integration
  • performance deterioration modeling

Technologies

NPSS GSP Python MATLAB pandas HDF5 Git

Work History

2025-01

Integrated engine performance model with ML surrogate for rapid trade studies. Provided 5,000 NPSS training points to ML team and validated surrogate accuracy against held-out test cases at on-design and off-design corners.

Challenge: Surrogate model accuracy degraded significantly at off-design corners — very low power settings and high-altitude windmill restarts. Required targeted additional sampling of the sparse design-space regions.

Learned: Engine performance surrogates need targeted sampling at off-design corners, not just uniform DoE grids. Physics-based knowledge of where model behavior changes should guide the sampling strategy.

NPSS Python pandas

2024-04

Engine performance deck database for integration with aircraft performance engineering tools. Generated 15,000 operating point calculations covering altitude, Mach, dISA, power setting, and bleed extraction combinations.

Challenge: NPSS is single-threaded — running 15,000 cases sequentially took 18 hours. Parallelized case generation using Python multiprocessing, reducing wall time to 2.5 hours.

Learned: NPSS batch execution benefits enormously from process-level parallelism. Each NPSS run is independent, so subprocess pool parallelism is straightforward and has near-linear scaling up to the core count.

NPSS Python pandas HDF5

2023-09

In-service performance deterioration model for the existing turbofan fleet. Correlated engine deck predictions with ACARS data from 120 aircraft over 3 years to quantify HPT and HPC wash-out rates.

Challenge: ACARS data quality was inconsistent — missing points, outliers from sensor faults, and different sampling rates across airlines. Spent 2 weeks on data cleaning before any analysis was possible.

Learned: For any in-service data analysis, budget at least 30% of total effort for data acquisition and cleaning. Clean data from 50 aircraft is more valuable than noisy data from 200.

GSP Python pandas

2023-03

Thermodynamic cycle study for a new turbofan variant — OPR and TET trade space analysis to maximize specific thrust while meeting SFC and NOx targets. Built NPSS model with variable cooling air bleed extraction.

Challenge: The OPR-TET trade space had discontinuities caused by the cooling air extraction model stepping between integer cooling row counts. Had to smooth the bleed fraction curve with a continuous interpolation function to enable gradient-based optimization.

Learned: Cooling air extraction models must be continuous for optimizer compatibility. Discrete cooling row assumptions create numerical artifacts that mislead optimization algorithms — always verify the objective function is smooth before running gradient-based search.

NPSS Python MATLAB