- Phase 0: Foundation & Documentation
- Phase 1: Mechanistic Engine (ALBATwin)
- Phase 2: Synthetic Data Generation
- Phase 3: Hybrid Modeling (PINN)
- Phase 4: User-Ready Product
Goal: Establish the theoretical and architectural groundwork.
- Requirement Analysis: Define PRD and User Stories.
- Architecture Design: Modular system design (Architecture.md).
- Mathematical Specification: Transcribe ALBA model equations (Math_Model.md).
- Environment Setup: Initialize
uvproject,pyproject.toml, and git repository.
Goal: Build a robust Python simulator that reproduces the results of Casagli et al. (2021).
- Core Implementation:
- Implement
ReactorConfigandStateVector. - Implement
BioKinetics(Stoichiometry + Rates). - Implement
HydroChemistry(pH solver).
- Implement
- Solver Integration: Connect components with
scipy.integrate. - Testing (TDD):
- Unit tests for individual kinetic functions.
- Mass balance verification tests (Critical).
- Validation: Compare simulation outputs with paper benchmarks (Figure 2 & 3 of Casagli 2021).
Goal: Use the validated engine to create a "Ground Truth" dataset for AI training.
- Scenario Definition: Define ranges for Light, Temperature, and Influent Loads.
- Batch Simulation: Run parallel simulations covering the operational space.
- Data Processing: Clean, normalize, and structure the output into training tensors.
- Noise Injection: Add Gaussian noise to simulate sensor inaccuracy (optional, for robustness testing).
Goal: Develop a Physics-Informed Neural Network that learns from data while respecting physical laws.
- Network Architecture: Design the neural network (Input: Env/State -> Output: dX/dt or Next State).
- Loss Function Engineering:
$L_{total} = L_{data} + \lambda \cdot L_{physics}$ - Implement
$L_{physics}$ using the residual of the Phase 1 differential equations.
- Training: Train the model using the synthetic dataset.
- Evaluation: Compare PINN performance vs. Pure Data-Driven models vs. Mechanistic Model.
Goal: Deploy the tool for end-users (researchers/operators).
- Dashboard: Build a Streamlit app for visualization and interaction.
- Dockerization: Create a
Dockerfilefor easy deployment. - Documentation: User manual and API reference.
- Publication: Prepare technical report or paper on the Hybrid Twin implementation.
Last updated: March 10, 2026 by Anibal Rojo