| English | 中文 |
Start with the LCA method: state goal, functional unit, product-system boundary, region, time, allocation and impact category. scripts/run_lca.py demonstrates only a linear single-output foreground LCI and LCIA calculation. It is not a full ISO 14040 study, a licensed inventory database, or a registered run-tool adapter.
Supply five CSVs in one input directory:
| File | Columns | Meaning |
|---|---|---|
processes.csv |
process,product,product_unit |
One unique reference product per process |
technosphere.csv |
product,process,amount |
Positive reference output, nonpositive input; list every product/process pair |
demand.csv |
product,amount |
Functional demand; list every product, using explicit zero where needed |
biosphere.csv |
flow,process,amount_kg |
Nonnegative elementary flow per process scale; list every flow/process pair |
characterization.csv |
impact_category,flow,factor,impact_unit |
One explicit factor per category and flow |
The script solves T s = d, computes g = B s, then sums characterized inventory score_k = Σ_f CF[k,f] g[f]. It rejects missing factors or matrix cells, duplicate identifiers, invalid sign conventions, ill-conditioned systems and negative process scales. It reports process_scales.csv, inventory.csv, impacts.csv and a hashed summary.json.
From the repository root, with NumPy installed, AIRMOSAIC_ASSETS_DIR set to an existing external asset root, and an unused output path:
$out = Join-Path $env:AIRMOSAIC_ASSETS_DIR 'results/lca-toy-example-v1'
python skills/model/01_socioeconomic_natural_change/life-cycle-assessment/scripts/run_lca.py --input-dir assets/data/01_socioeconomic_natural_change/SyntheticLCA/v1 --out-dir $out
The manifest-listed synthetic LCA demands one toy product unit, requires 0.5 kg toy material and yields 2 kg CO2e for its single climate_change indicator. Verify this against the output and toy-manifest.json. The script cannot model co-products, recycling allocation, dynamic emissions, uncertainty distributions or multiple locations. For real datasets and broader LCIA, Brightway is an open-source scripting path; database access, method selection and scientific review remain separate. Keep research data and outputs under authorized external asset roots.
The public Tiangong LCA skills informed the separate inventory-review and calculation gates, while the K-Dense industrial-ecologist agent skill informed boundary checks. Their runtime and datasets are not part of this script.