Cyber-physical AI – AI that interacts with machines and processes under control

What makes AI cyber-physical

Cyber-physical systems connect the physical world – machines, controllers, sensors – with software that evaluates data and prepares decisions. Cyber-physical AI adds AI models that understand not only text, but process values, time series and the relationships between plant and business process. The term physical AI is also used for this.

Capture, evaluate, act under control

OPAIRS builds cyber-physical AI in three stages. The first two are in use today; the third we are developing together with our customers.

• •Capture: machine, process and business data in one pipeline
• •Evaluate: specialised models with measurable quality per use case
• •Act under control (in development): guardrails and approvals built in

Connecting machines on-premise

With OPAIRS Connect Machine, the system reads data from OPC UA, Siemens S7, Beckhoff TwinCAT, Modbus, EtherNet/IP and ifm IO-Link and stores it in a local database. Data flows in one direction, from the machine into the system – the module does not write anything back to controllers. Everything runs in-house, without the cloud.

Why not just an AI agent?

An agent that acts without context and without limits is a risk in production. Cyber-physical AI is bound to plant, process and approvals: it knows the data situation, works within defined limits and makes the basis of its decisions traceable. This turns AI into a controllable part of manufacturing.

Towards the self-learning machine

The goal we pursue together with our customers: a machine that learns from its own operation. Today, approvals and results are already captured and models are tuned to specific use cases. Step by step, this leads to processes that run reliably and automatically – with the compute power right at the customer’s site.