Applied Computing has unveiled the next generation of its artificial intelligence platform, Orbital 2.0, introducing what the company describes as the first-ever world model for energy that understands cause and effect across an entire industrial plant.
Launched at ADIPEC 2026 in Abu Dhabi on November 4, the upgraded platform marks a shift from conventional AI systems that identify patterns in operational data to a physics-based model designed to understand how an energy facility actually functions.
The new architecture enables Orbital to learn how mass and heat move through interconnected equipment and how these interactions influence overall plant performance. This, the company said, allows operators to assess the wider consequences of operational decisions before implementing them, while seeking to increase production without compromising plant reliability and safety.
Unlike the earlier version, which relied on identifying patterns in plant data to execute predefined tasks, Orbital 2.0 is designed to adapt to different operational requirements and engineering workflows.
The platform is built on a knowledge graph representing the physical interactions within a plant. Applied Computing said each customer will have an independent world model operating within its own environment, ensuring that customer data and models remain private.
Orbital 2.0 also introduces a set of features aimed at giving engineers greater control over plant analytics and automation.
These include automatically generated dashboards based on engineers' requirements or suggested configurations, with information drawn directly from live plant data. The platform can also convert recurring operational activities, including daily shift reports and heat and mass balance checks, into automated applications through simple prompts.
Another key addition is recursive self-learning, which allows the system to improve through interactions with engineers and feedback provided during operations. The approach is intended to retain critical plant knowledge within the system, even when experienced personnel move on.
The company said the new capabilities eliminate the need for its researchers to manually tune and adapt models within individual customer environments. This is expected to reduce deployment cycles from weeks to days, while enabling engineers to customise and extend the platform themselves.
Callum Adamson, CEO and co-founder of Applied Computing, said: “Our mission is to get every energy system running at its physical optimum, and the fastest way to do that is to put Orbital in engineers’ hands. Until now, adapting the model to each plant relied on our researchers. That barrier is gone: Orbital now learns for itself and can be fully customised by the engineers who use it every day. We are bringing it to ADIPEC because this is where the industry comes to decide what the next decade of energy looks like.”
Samyakh Tukra, Chief AI Officer and co-founder of Applied Computing, said: “The first version of Orbital learned the patterns in a plant’s data. What we have built now is a world model. It learns cause and effect, understands how a plant actually behaves, and gets better every day it runs. A plant is always evolving, and so is the work of the engineers who run it, so Orbital cannot stand still either.”
Orbital is designed for complex energy and industrial facilities, including refineries, petrochemical plants, LNG facilities, ammonia units, and olefins plants. It connects with live and historical plant data to identify operational problems, explain their underlying causes using process physics and recommend corrective actions.
Applied Computing said the technology is already being deployed at some of the world's largest energy and petrochemical assets. Through KBR INSITE, the platform is also being used on ammonia and olefins units.
