Hybrid Models

An engineering module for building hybrid models and physics-informed neural networks (PINNs) to predict process variables, analyze and optimize operating conditions

Physics and data

Combining material and heat balances, engineering relationships and machine learning models to model equipment behavior and refine relationships using data

Physics-informed neural networks

PINNs incorporate process equations, conservation laws, initial and boundary conditions into neural network training to solve forward and inverse problems

Model training and validation

Training on plant, experimental and simulation data, validation against independent datasets, and assessment of accuracy and the domain of applicability