spacedc-mdao 0.4.0

Multidisciplinary design analysis and optimization of orbital compute infrastructure, with terrestrial data-center baselines.

spacedc-mdao

A Python package for multidisciplinary design analysis and optimization (MDAO) of orbital compute infrastructure, with terrestrial data-center baselines for comparison.

The package optimizes delivered useful compute, not nominal watts or nominal GPUs. It takes installed capacity and degrades it through power, thermal, network, reliability, and utilization limits, then reports where a design fails and which assumptions decide the outcome. Its job is to make the feasibility boundary visible, not to argue that space wins.

Documentation: https://jman4162.github.io/spacedc-mdao/ (quick start, user tiers, model architecture, API reference, and a generated assumptions/provenance table). See SPEC.md for the design contract and background_information/EQUATIONS.md for the governing equations.

Status

Phases 1 and 2 are complete. Implemented:

• Core (Phase 1): scenario loading, the discipline models in the v0.1 minimum set (background_information/EQUATIONS.md §15), the delivered-compute waterfall, an Earth-vs-space comparison with binding-constraint diagnosis and feasibility thresholds, Monte Carlo and tornado sensitivity, matplotlib plots, and a CLI.
• Thermal radiator co-design (2A): a thermal/ module coupling chip power → junction temperature → coolant loop → radiator temperature → area → mass, with net radiator flux (emission minus absorbed solar/albedo/Earth-IR at end-of-life coatings), a bottleneck classifier, and validation anchors (ISS PVR/EATCS, Starcloud, NASA high-temp). The radiator temperature is bounded by the chip stack; closure is checked at end of life.
• MDAO + optimization (2B): OpenMDAO components wrap the evaluator (FD partials), optimize_single for constrained single-objective optimization, pymoo NSGA-II Pareto fronts, a scipy Latin-hypercube DOE, and SALib Sobol indices.
• Interactive dashboard (2C): plotly figures (delivered/cost/mass/power-sankey, tornado, Pareto, constellation graph, thermal panels) and a Panel app, plus an assumption-provenance table.
• Fidelity (2D): environmental CO₂e/water accounting, station-keeping Δv and propellant, beta-angle eclipse, an optional Skyfield ground-access seam, and five Earth baselines.
• Credibility (3A): all soft factors moved to provenance-tagged catalogs; Starcloud/ISS recovered from the model; orbit-dependent radiation (TID/SEU); RF margin and optical weather availability bind; launch-cost cases; input validation.
• Real MDAO (3B): mixed-integer architecture optimization (satellites, accelerators/sat, altitude); transient orbit thermal; a workload library (space-native vs Earth-dependent); uncertainty fan / orbit-timeline / link-budget-heatmap figures.
• UX (3C): CLI provenance/doe/sobol/--version, HTML/Markdown report export, Evaluation.to_dict(), and a beginner→advanced notebook ladder.

The package is skeptical by default. For the bundled 1 MW inference scenario, Earth wins on levelized cost: the orbital design is downlink-limited, its radiators are a multi-tonne burden (chip-limited temperature), and station-keeping plus replacement add up.

Install

spacedc-mdao uses uv. The base install is light; capabilities live behind extras.

uv sync                                          # base (numpy/scipy/pydantic/pint/matplotlib)
uv sync --extra dev --extra mdao --extra viz     # development + optimization + dashboard

Extra	Pulls in	Enables
mdao	openmdao, pymoo, SALib	optimization, Pareto fronts, DOE, Sobol
viz	plotly, panel, networkx	interactive figures + dashboard
orbit	skyfield	ground-station access windows (needs ephemeris)
rf	opensatcom	Tier-1 RF link-budget backend (else inline Friis/FSPL)

Use

import orbitdc as odc

space = odc.load_scenario("examples/scenarios/orbital_1mw_inference.yaml")
earth = odc.load_scenario("examples/scenarios/earth_hyperscale_baseline.yaml")

result = odc.compare(space, earth)
print(result.summary())
print(result.explain_binding_constraints())

From the command line:

orbitdc compare  examples/scenarios/orbital_1mw_inference.yaml examples/scenarios/earth_hyperscale_baseline.yaml
orbitdc optimize examples/scenarios/orbital_1mw_inference.yaml --pareto lcoc,kg_per_kw      # needs [mdao]

Notebooks: examples/notebooks/01_compare.ipynb (Earth-vs-space) and examples/notebooks/02_radiator_feasibility.ipynb (how big and heavy the radiators must be). Dashboard: uv run panel serve examples/dashboard_app.py --show (needs [viz]).

Development

uv run ruff check . && uv run ruff format --check .
uv run mypy src
uv run pytest

All code must pass ruff, ruff format, and mypy --strict before commit. CI enforces this. Every default number in the catalogs carries provenance (source, date, confidence, kind); see background_information/THEMRAL_RADIATOR_DEEPDIVE.md for the thermal modeling background.

License

MIT.