Deploy from GitHub Actions/CLAUDE.md

Tennessee Eastman Process Simulator

Project Overview

Python interface to the Tennessee Eastman Process (TEP) simulator. The TEP is an industrial chemical process benchmark for control systems and fault detection research. Includes both a pure Python implementation and optional Fortran acceleration.

Build & Install

# Default install (Python backend only, no compiler needed)
pip install -e .

# With Fortran acceleration (requires gfortran)
pip install -e . --config-settings=setup-args=-Dfortran=enabled

# With optional dependencies
pip install -e ".[dev]"        # Dev tools (pytest, matplotlib, dash)
pip install -e ".[web]"        # Dash web dashboard

Backend Selection

from tep import TEPSimulator, get_available_backends, is_fortran_available

# Check available backends
print(get_available_backends())  # ['python'] or ['fortran', 'python']

# Use specific backend
sim = TEPSimulator(backend='python')   # Pure Python (always available)
sim = TEPSimulator(backend='fortran')  # Fortran (if installed)

Test Commands

pytest                           # Run all tests
pytest tests/test_simulator.py   # Run specific test file
pytest -xvs tests/test_simulator.py::test_name  # Run single test with output

CLI Commands

tep-sim --duration 2 --faults 1 --output data.dat  # Batch simulation
tep-web                                             # Launch Dash dashboard

Architecture

• tep/simulator.py - High-level TEPSimulator interface (backend-agnostic)
• tep/python_backend.py - Pure Python implementation of TEP process
• tep/fortran_backend.py - f2py wrapper for Fortran TEINIT/TEFUNC (optional)
• tep/constants.py - Physical constants, initial states, variable names
• tep/controllers.py - PI controllers, decentralized control
• tep/controller_base.py, controller_plugins.py - Controller plugin system
• tep/detector_base.py, detector_plugins.py - Fault detection framework
• tep/cli.py - Batch simulation CLI (tep-sim)
• tep/dashboard_dash.py - Dash web dashboard (tep-web)
• tep/_fortran/ - Compiled Fortran extension (optional)

Key Patterns

• Python backend is default; Fortran is optional for ~5-10x speedup
• Fault detectors use plugin system with @register_detector decorator
• Controllers use plugin system with @register_controller decorator
• SimulationResult dataclass holds time, states, measurements, mvs arrays
• Both backends produce statistically similar results (<1.5% difference)

Testing Notes

• Tests compare Python outputs against reference .dat files
• Random seed control enables reproducibility within each backend
• Use pytest -xvs for verbose output when debugging