Industrial AI glossary.

Agentic AI

AI systems that plan and carry out multi-step tasks, such as gathering data, running checks and drafting a recommendation, rather than only answering a single prompt. In industrial settings, agents must work inside defined rules and permissions and hand decisions to a human. Atelic deploys governed agents on top of its knowledge graph.

CMMS (computerised maintenance management system)

Software that records assets, work orders, maintenance history and spare parts, for example SAP PM or IBM Maximo. CMMS history is one of the richest sources for reliability and failure analysis.

Data historian

A time-series database that stores high-frequency process data (pressures, temperatures, flows, power output) from control systems, often for years. Historians such as AVEVA PI hold the operating record that AI models learn from.

Data sovereignty

The principle that data stays under the control, and within the legal jurisdiction, of the organisation that owns it. For critical infrastructure this usually means private-cloud or on-premise deployment with no data leaving the operator's boundary.

DCS (distributed control system)

A plant-wide control system that runs continuous processes such as refineries, gas plants and power stations through networked controllers and operator consoles.

Deterministic AI

AI whose outputs follow explicit rules, physics and constraints, so the same inputs give the same, explainable result. It contrasts with purely probabilistic text generation, and is preferred where decisions affect safety, compliance or large capital spend.

Digital twin

A live digital model of a physical asset, process or site that is kept in sync with operating data, used to monitor performance and test scenarios before acting on the real asset.

Flaring

The controlled burning of natural gas at oil and gas facilities, usually for safety or when gas cannot be captured or sold. Reducing routine flaring cuts both emissions and lost product.

Heat rate

The fuel energy a thermal power plant uses to produce one kilowatt-hour of electricity, typically in Btu/kWh or kJ/kWh. A lower heat rate means higher efficiency; a rising heat rate is an early sign of equipment degradation.

Human in the loop

A design in which AI proposes and a qualified person reviews, approves or overrides before action is taken. It keeps accountability with engineers and operators.

Industrial AI

Artificial intelligence applied to physical operations such as oil and gas, power and manufacturing. Unlike consumer AI, it must handle sensor data and engineering documents, respect physics and safety limits, and give answers that can be audited.

Industrial DataOps

The practice of collecting, cleaning, contextualising and delivering industrial data to the people and applications that need it, in a governed and repeatable way.

Industrial knowledge graph

A connected model of assets, tags, documents, people and events, and the relationships between them. It lets software answer questions such as "which pumps share this seal type and have failed twice this year?" across sources that were previously siloed. It is the core of the Atelic Knowledge System.

IT/OT convergence

Bringing together information technology (ERP, finance, documents) and operational technology (SCADA, DCS, PLCs, historians) so business and operating data can be analysed together.

MES (manufacturing execution system)

Software that manages and records production on the shop floor: orders, batches, materials, quality checks and equipment status. It sits between the ERP and plant control systems.

P&ID (piping and instrumentation diagram)

An engineering drawing that shows a process plant's piping, equipment, instruments and control loops. P&IDs are a key source for linking sensor tags to physical equipment.

Pipeline integrity management

The programme of inspection, monitoring, risk assessment and repair that keeps pipelines operating safely and prevents leaks and failures.

PLC (programmable logic controller)

A rugged industrial computer that controls machines and process steps in real time, reading sensors and driving valves, motors and drives.

Predictive maintenance

Using operating data and models to forecast when equipment is likely to fail, so maintenance is planned before a breakdown rather than on a fixed schedule or after failure.

SCADA (supervisory control and data acquisition)

A system that monitors and controls geographically spread assets such as pipelines, well sites, substations and grids, collecting real-time data from field devices.

Institutional knowledge (tribal knowledge)

Expertise held by experienced personnel rather than in documented systems. Capturing it before engineers retire is a primary reason operators adopt knowledge management platforms.

Unified namespace (UNS)

A single, structured, real-time data layer where every machine, system and site publishes its data in a common model, usually organised by enterprise, site, area and line. Applications read from the UNS instead of building point-to-point integrations.

Require a further definition? Ask our team or see the FAQ.