moonbit-reactor

moonbit-reactor is a deterministic MoonBit toolkit for early-stage chemical reactor design. It provides typed CSTR, PFR, and batch models, reaction kinetics, thermal screening, reactor-train analysis, numerical solvers, sensitivity studies, uncertainty intervals, engineering metrics, and offline benchmark reports.

Core capabilities

• Isothermal, adiabatic, and lumped jacketed CSTR/PFR/batch calculations.
• Zero-, first-, and second-order kinetics, Arrhenius correction, reversible, series, and parallel helpers.
• Bisection diagnostics, trapezoid/Simpson/adaptive quadrature, Euler and RK4 integration.
• Axial and batch profiles, reactor-train analysis, conversion/selectivity/productivity metrics.
• Deterministic sensitivity scans, interval propagation, uncertainty summaries, and bounded control helpers.
• Design envelopes, grid-search optimization, cost screening, validation reports, Markdown and CSV output.
• Offline benchmark cases with source URLs, units, assumptions, expected values, and tolerances.

Quick start

moon run cmd/main
moon run cmd/benchmarks

Library example:

///|
test {
  let reaction = Reaction::new(name="A -> B", order=First, k_ref=0.5)
  let feed = Feed::new(concentration=1.0, volumetric_flow=2.0, temperature=300.0)
  let point = design_pfr(reaction, feed, 4.0)
  assert_true(point.conversion > 0.63 && point.conversion < 0.64)
}

CLI

• moon run cmd/main runs a compact engineering example.
• moon run cmd/benchmarks executes all offline regression cases and prints Markdown and CSV-ready results.

Architecture

The root package owns the public data types and design APIs. Focused modules separate kinetics, numerical methods, profiles, networks, validation, reporting, optimization, thermal properties, sensitivity, control, and safety screening. cmd/main and cmd/benchmarks consume the same public APIs as downstream packages.

See docs/DESIGN.md for assumptions and docs/benchmarks.md for provenance.

Benchmarks

The benchmark suite is offline and reproducible. Analytical cases use closed-form reactor identities; source-backed records preserve their units and assumptions; screening cases are explicitly labeled as engineering scenarios. Run:

moon run cmd/benchmarks

Testing and CI

moon fmt --check
moon check --deny-warn --diagnostic-limit 300
moon test --deny-warn --diagnostic-limit 300
moon test --deny-warn --target native --diagnostic-limit 300
moon info
git diff --check

GitHub Actions runs formatting, strict checks, interface generation, tests on native and default targets, and the offline benchmark command across Linux, macOS, and Windows.

Scope

This package targets deterministic engineering education and early process screening. It does not claim CFD, multiphase simulation, thermodynamic property prediction, unit algebra, plant safety certification, or arbitrary reaction-graph solving.

License

Apache-2.0. See LICENSE.