The problem nobody puts on the slide
Every conversation about AI in energy eventually arrives at the same place. Before a model can say anything useful about a plant, someone has to find the data, make sense of it and get it out safely. On most sites that work is slow, manual and never quite finished.
A typical unit describes itself in several systems that were never designed to talk to each other. Sensor history sits in the historian. Product quality sits in the lab system. The design lives in drawings and datasheets. Maintenance history sits in the ERP. Shift logs and operating procedures are in documents, some of them scanned. Each system is reliable on its own terms, and almost none of them are joined up.
So the knowledge exists, but it is scattered. Engineers fill the gaps with memory and experience, which works until the person who remembers moves on.
Three things that make plant data hard
It is scattered. The same piece of equipment appears under different names in different systems, and nobody owns the map between them.
It is inconsistent. Tags change names, units differ, clocks drift, instruments go out of calibration and gaps appear when a historian was upgraded years ago.
It is hard to reach safely. Operational systems sit behind strict security for good reason. Getting data out must never put the plant at risk.
Any AI project that skips these problems ends up with a model trained on data that does not mean what it appears to mean. That is how pilots stall.
What Orbital Data Connect does
Orbital Data Connect is the foundation layer of Orbital. It brings the plant's data together in one governed place, cleaned and aligned, without replacing any of the systems a site already runs.
It does four jobs. It discovers what data exists and where. It brings that data across through a read-only link. It aligns it, so the same equipment, the same units and the same time base are used throughout. And it reads documents and drawings, so the context engineers rely on sits alongside the numbers.
The result is a single, consistent picture of the unit that Orbital, and the customer's own teams, can work from.
Read-only, by design
Data Connect only reads. It never writes to the control system, never changes a set-point and never touches alarms or safety systems. Data leaves the operational network through a secure gateway behind the customer's firewall, using encrypted transfer and the open industrial standards plants already use, such as OPC UA, MQTT and REST.
This matters for more than security reviews. It means a site can connect Orbital without a change to how the plant is run, which shortens the path from first conversation to first useful answer.
Governed and traceable
Data lands in a lakehouse the customer controls, built on Databricks, with full lineage. Every number Orbital shows can be traced back to the original tag, sample or document it came from. If an engineer asks where a figure came from, the answer is one click away.
If an engineer cannot see where a number came from, they will not act on it. Traceability is not a compliance feature. It is what makes the answer usable.
Useful beyond Orbital
Because the data is cleaned, aligned and stored in the customer's own environment, it is not locked inside a product. The same governed data can serve reporting, analytics and data science teams. Many sites find that this alone is worth the effort, because it is the first time the whole unit's history has been in one place.
What changes for the team
Engineers stop spending days pulling data from four systems before they can start on the actual question.
Answers come with their sources, so reviews move faster and arguments about whose number is right go away.
New starters can find what they need without knowing which system or which person to ask.
Every other part of Orbital, from forecasting to optimisation, starts from the same trusted foundation.
Where to start
We start with one unit and one question. Data Connect brings in the history and documents for that unit, and within weeks the team can see what Orbital learns from its own plant. From there the same foundation extends unit by unit.
Book a technical briefing to talk through your own systems.
About the author
Orbital product team
Applied Computing
The engineers, data specialists and AI researchers who design and build Orbital, working alongside the operators who use it.
