# Figure sources and reproduction notes

## Historical NASA material

Figures A and B are cropped renderings, not newly drawn or generated diagrams:

- **nasa-engine-schematic.png**: Figure 1, printed/PDF page 2, “Simplified diagram of engine simulated in C-MAPSS”.
- **nasa-simulation-modules.png**: Figure 2, printed/PDF page 2, module connections in the simulation.

Source: A. Saxena, K. Goebel, D. Simon and N. Eklund (2008), *Damage Propagation Modeling for Aircraft Engine Run-to-Failure Simulation*, PHM 2008. [Official NASA-hosted paper](https://c3.ndc.nasa.gov/dashlink/static/media/publication/2008_IEEEPHM_CMAPPSDamagePropagation.pdf). The paper itself attributes these figures to reference [11], D. K. Frederick, J. A. DeCastro and J. S. Litt, *User's Guide for the Commercial Modular Aero-Propulsion System Simulation (C-MAPSS)*, NASA/TM-2007-215026.

[NASA NTRS record 20090029214](https://ntrs.nasa.gov/citations/20090029214) lists the alternate title *Damage Propagation Modeling for Aircraft Engine Prognostics*, the same authors and 2008 PHM meeting, public distribution, and “Public Use Permitted.” Its direct PDF endpoint returned 403 during this audit; the official NASA Dashlink PDF above was accessible. No claim of NASA affiliation or endorsement is made.

Extraction: PyMuPDF, PDF page index 1, four pixels per PDF point, grayscale. Figure 1 is converted from the source's restrained color drawing to grayscale; Figure 2 is black and white. No technical labels or connections were redrawn. Source English labels remain unchanged in both site languages. The source PDF is not redistributed in this repository.

Crop rectangles in PDF points (x0, y0, x1, y1):

- Figure 1: (365, 452, 526, 558)
- Figure 2: (311, 577, 534, 663)

Source PDF SHA-256: `e7aaef80c177333f400a4c1099fe76e67d244569b9397c937ec4cd3ed5b44a27`.

No manual figure placement remains necessary. If replacing these crops, use the identified original figures and retain their source attribution.

## Project evidence

- `prediction-evidence-en.png` / `-de.png`: generated from `docs/evidence/predictions.csv`. All points are saved-model inference results with the explicitly reconstructed scaler; no fabricated or fitted-to-test predictions.
- `degradation-en.png` / `-de.png`: raw FD001 training sensor s_11, engines 1–3; capped and uncapped RUL for engine 1. Illustrates data and target construction, not prediction performance.
- `training-history-en.png` / `-de.png`: original `models/training_log.csv`; not a new training run.
- `original-test-result.png`: byte-for-byte copy of `models/test_result_plot.png`. Historical RMSE annotation and “Accuracy Analysis” wording are retained only in this archived image. The original plot is linked separately from newly generated evidence.
- `logo-turbofan.svg`: original project identity, simple front-view fan line art. Decorative, not an engineering schematic or NASA insignia.

Regenerate the six evidence plots and request payload from the repository root:

```sh
MPLCONFIGDIR=/tmp/turbofan-mpl python scripts/generate_report_figures.py
```

Prerequisites: exported predictions, pandas, NumPy, matplotlib and scikit-learn. Figure generation never trains or modifies the saved model.