GymChem

English | 中文

Gymnasium-compatible simulation environments for chemical process optimization and data-driven analysis — from a minimal CSTR to industrial-scale benchmarks.

Why GymChem

Chemical process optimization research is fragmented: pyTEP is a simulator only, IDAES is powerful but steep, and ChemGymRL targets lab operations rather than industrial processes. GymChem unifies these scenarios behind a standard Gymnasium API — consistent state/action/reward design, fully traceable parameters, reproducible baselines and leaderboards, plus datasets and data-analysis tutorials. For researchers it is a ready-to-use benchmark platform; for students it is a friendly on-ramp to RL for chemical engineering.

Environments

CSTREnv models a continuous stirred-tank reactor with a first-order exothermic reaction A → B: species balance, energy balance with Arrhenius kinetics, a cooling-water jacket, and a conservation-checked explicit-Euler integrator.

Install

# from source
git clone https://github.com/guyuan0710/gymchem.git
cd gymchem
pip install -e ".[dev]"

# once published on PyPI
pip install gymchem

Quick start

from gymchem.envs.cstr import CSTREnv, CSTRSimpleEnv

# v0.1: exothermic A→B CSTR — temperature + concentration + jacket
env = CSTREnv()
obs, info = env.reset(seed=0)
for _ in range(100):
    obs, reward, terminated, truncated, info = env.step(env.action_space.sample())
    if terminated or truncated:
        break

# v0.0: minimal temperature tracking
env = CSTRSimpleEnv()
obs, info = env.reset(seed=0)

Test

pip install -e ".[dev]"
pytest        # 17 tests: env_checker, no-NaN runs, conservation residuals, data pipeline
ruff check .  # zero warnings

Data validation

Automated detection, validation, and update-maintenance for research/experimental data (gymchem.data, no new dependencies):

python -m gymchem.data data/demo_cstr_trajectory.csv \
  --manifest data/manifest.json --sidecar data/demo_meta.json \
  --report data/demo_report.md

See docs/data-validation.md for the full workflow.

Docs & examples

• Parameter provenance (CSTR) — every physical constant traced to a source
• Data validation (detect / verify / maintain)
• Colab-ready notebook
• 中文说明

License

MIT © GymChem contributors