Data and Code for: Hybrid Modelling of Chemical Processes - A Unified Framework

This dataset contains the Python source code and synthetic data required to reproduce the results in the paper, "Hybrid Modelling of Chemical Processes: A Unified Framework Based on Deductive, Inductive, and Abductive Inference."

The project implements a layered hybrid modeling framework for a batch polymerization reactor, combining physics-based models with data-driven methods. The framework is composed of three distinct layers:
- Deductive Layer (Tp): Enforces first-principles mass and energy balances to simulate the physical dynamics of the reactor.
- Inductive Layer (Tm): An LSTM-based neural network that learns the unknown reaction kinetics from process data.
- Abductive Layer (Ta): A feedforward neural network that functions as a soft sensor to infer latent (unmeasured) variables such as molecular weight, viscosity, and branching index.

The dataset includes all necessary Python scripts to generate synthetic data, define and train the neural network models, run the integrated hybrid simulation, and visualize the results. The framework is built using Python with libraries including PyTorch, SciPy, and Scikit-learn.