Digital Twins
Physics-based digital twin technology explained.
SUPi's digital twins are not static 3D models or simple dashboards. They are physics-based virtual replicas that simulate real-time stress, fatigue, thermal behavior, and degradation based on actual operating conditions.
Why Physics-Based?
Pure data-driven models break when conditions change. Physics-based twins understand why equipment behaves the way it does, making predictions dramatically more accurate in complex environments like offshore platforms, chemical plants, and wind farms.
Supported Twin Types
| Asset Type | Physics Models |
|---|---|
| Rotating Equipment | Vibration dynamics, bearing wear, shaft fatigue |
| Heat Exchangers | Fouling progression, thermal efficiency, tube degradation |
| Pipelines | Corrosion modeling, fatigue cycling, pressure loss |
| Reactors | Catalyst deactivation, thermal runaway, yield modeling |
| Turbines | Blade stress, pitch optimization, power curve |
Hybrid Modeling
SUPi combines first-principles engineering knowledge with machine learning:
- Physics layer — encodes known thermodynamic, mechanical, and chemical relationships
- ML layer — learns plant-specific patterns and compensates for unmeasured variables
- Calibration — continuous Bayesian updating as new data arrives
# Digital twin definition
twin:
asset_id: COMP-001
type: centrifugal_compressor
physics_model: rotating_machinery_v3
ml_augmentation: true
calibration:
method: bayesian_online
update_frequency: "1h"
degradation_modes:
- bearing_wear
- seal_leakage
- impeller_fouling