Released test suite
33 tests passed across core fitting, projection, CLI behavior, data loaders, long-form validation, and the executable notebook.
Analysis software · Analysis software
Battery Cycle-Life Analyzer
Fit capacity fade, bound the forecast, and show when the evidence runs out.
Released Python software compares linear, power-law, and logarithmic fits, selects by RMSE, and returns no EOL estimate when a threshold is outside the three-times-observed projection window.
Engineering question
Which empirical degradation model best describes observed capacity fade, and how far can EOL projection be bounded honestly?
Fit three transparent empirical degradation families, compare fit quality, and return a bounded EOL or no estimate when the threshold is unsupported.
System and evidence boundary
Provenance enters before fitting—not after the chart.
The long-form path preserves cell and protocol context, while the legacy CSV/TSV path remains compact for cycle-capacity tables.
Data-provenance pipeline
- 01CSV, TSV, or long formCycle + capacity + optional protocol context
- 02Validation envelopeFinite values, bounds, cell ranges, timestamps
- 03Three empirical fitsLinear, power-law, logarithmic
- 04Model selectionLowest released RMSE criterion
- 05Bounded EOL / RULThreshold search stops at 3× observed cycles
- 06Publication figureAxes, units, label, provenance, source
Verified results
Figures keep scenario, units, source, and evidence class attached.
Every visual below points to its source data or retained source format and to the release- or commit-specific implementation behind it.
Three-model fit comparison
Generated with v0.1.0 using synthetic_lfp(cycles=1500, seed=42, noise_std=0.005). The right panel reports released RMSE and R² calculations.
- Provenance
- v0.1.0 · bf62fff47e3e · verified 2026-07-28
- Reproduce
python scripts/generate-open-engineering-figures.py --sources-root <audit-root>
Selected model, EOL threshold, and RUL
The observed-cycle boundary and full 3× search window are visible. If no crossing exists in that window, the released API returns None instead of extending indefinitely.
- Provenance
- v0.1.0 · bf62fff47e3e · verified 2026-07-28
- Reproduce
python scripts/generate-open-engineering-figures.py --sources-root <audit-root>
Relative Arrhenius acceleration
Activation energy is fixed at 0.5 eV and 25 °C is the reference. The utility is independent of the fitted capacity-fade trajectory and does not validate a lifetime model.
- Provenance
- v0.1.0 · bf62fff47e3e · verified 2026-07-28
- Reproduce
python scripts/generate-open-engineering-figures.py --sources-root <audit-root>
Validation and evidence
What passed, what was demonstrated, and what remains unvalidated.
Thirty-three package, CLI, data-schema, and notebook tests passed at the released revision.
Synthetic disclosure
Bundled LFP and NMC datasets are demonstrations, not measurements. Every portfolio plot on this page is labeled synthetic or illustrative.
Projection behavior
EOL is searched only to three times the largest observed cycle. Unsupported crossings return no estimate.
Uncertainty gap
v0.1.0 does not calculate parameter confidence intervals or propagate measurement uncertainty into RUL.
Shortest verified path
Pin the audited source before running it.
git clone https://github.com/mohammadrezwankhan/battery-cycle-life-analyzer.git
cd battery-cycle-life-analyzer
git checkout v0.1.0
python -m pip install -e ".[dev]"
python -m bcla --model all
python -m pytest -qVerification environment: Python 3.14.6 with NumPy, SciPy, and Matplotlib on Windows 11. Distinguish running the scientific model from rebuilding site figures (which uses scripts/generate-open-engineering-figures.py in the portfolio website repository). Commands are concise entry points; repository documentation remains authoritative for setup details.
Documentation map
Go directly to the useful evidence.
Engineering boundary
The artifact cannot qualify the decision by itself.
- The models are empirical curve fits, not electrochemical, physics-of-failure, warranty, or safety models.
- EOL and RUL depend on representative laboratory data, protocol, temperature, operating window, model family, and extrapolation distance.
- The Arrhenius comparison is a separate relative sensitivity; it is not coupled to capacity fade and cannot establish absolute life.
- Use measured project data and uncertainty analysis before using any projection in an engineering, warranty, or asset-management decision.
Related professional context
Follow the evidence into adjacent work.
Use the pinned source, inspect the evidence, and keep project authority with a qualified human reviewer.
Open v0.1.0
Return to open engineering
Evidence summary
Reviewed 2026-11-12; evidence records retain their own scope and dates.
- Documentation snapshot
- 2026-10-06 · 65df747 ·
65df747ccf1ff0d2427ab8ba321b4c6839d0ad3b - Portfolio source record
- 33 results · executed-upstream · bf62fff ·
bf62fff47e3e8759622c346bd66ae484e0c0ee8c - Current release
- v0.2.0 · published 2026-08-27 · tag target
4ac3869071e0e0f28cc701a6b2bbb8fe369a2a97 - Current verification
- Successful hosted run · 2026-11-10 · target
78d3de863ca65de7fe5a28911eb8167e7db9e2c4 - Repository head at 2026-10-17 inventory
630c030c806c81523554b2c40085e13424e0ca9d· inventory snapshot only