Is a digital representation of the design and operational performance of a physical equipment in its real state.

The physical design of components and their designed-in interactions inside an equipment, its operating life history, maintenance records, unique operating conditions are all “model metadata” that go into creating this unique representation of a digital twin.

Additionally, the model updates itself with every change in operational parameters of its underlying physical twin. In other words, the digital twin lives its life as an accurate proxy of the physical self, and lends itself to  interrogation for purposes of analysis of its physical state, for performance simulations and for prediction of possible failures.

A base infrastructure required to make the Digital Twin a reality involves a wide array of sensors to capture performance data from each component and transmit them to the data-management platform for transformation and storage.

Sensors collect data relating to process cycles, temperature & pressure, flow-rates, wear and tear, oil-pressure variations from one part of a machine to another, the torque of connected tools during tightening operations, or the status of the individual parts in a hydraulic valve. This flood of data is a new raw material for connected industry.

The other important piece is the availability of machine learning algorithms customised for ingestion of large volumes of equipment data and finding patterns across millions and billions of time-value pairs to infer a “black-box” model that codifies complex statistical relationships among hundreds (or thousands) of operating parameters. With good algorithms and the right software, it can shed light on new correlations.

Such models allows for projection of future performance and signal probabilities of potential catastrophic equipment failure down the line. The signal becomes a critical decision input for minor unscheduled repair or major scheduled shutdowns.