Force-X: Hybrid Digital Twin
Physics-Based Modeling + Neural Networks
Force-X is the fusion of physics-based modeling and neural network-based deep machine learning. This makes the thermodynamics, kinetics, and hydrodynamics fully embedded in the machine learning model. Hence, Force-X provides unmatched predictive insight and asset optimization capabilities.
Hybrid Digital Twin Generation Workflow
(80% of Datapoints)
Why Force-X?
Limitations of Standard ML Twins
- • Tailored for discrete event processes: Hydrocarbon processing are continuous operations.
- • Near-Infinite Plant Data Required: ML models need enormous training sets to map all relevant variables.
- • Mass Balance Errors (2-5%): Standard neural nets do not enforce physical conservation laws, leading to physically impossible outputs.
- • Poor Non-Linear Mapping: Struggles to accurately represent highly non-linear chemical and thermal boundaries.
- • Limited Accuracy: Overall accuracy typically hovers only between 80% to 90%.
Force-X Hybrid Advantage
- • No AI Hallucinations: Neural outputs are hard-constrained by physical equations.
- • Zero Mass Balance Deviations: Mass, energy, and momentum balances are conserved dynamically.
- • Physics-Guided Machine Learning: Incorporates unit operations theory directly into DNN algorithms.
Try Force-X In Real Time
Force-X Simulation Portal
Enter temporary account credentials below
Physically Constrained Neural Networks
Our interactive simulator environment allows engineers to test how first-principles mass, energy, and momentum equations are hard-constrained inside deep learning models.
By running catalytic reforming cases or column distillation bounds, users can verify that error tolerances consistently achieve R² > 0.95 while eliminating physical anomalies.