When AI Understands the Factory: A New Era in Industry

[Erkan Teskancan]

​

## When AI Understands the Factory: A New Era in Industry

Operational technologies (OT) are at a critical turning point. In recent years, process manufacturers have been monitoring and controlling more areas of the factory than ever before, thanks to decreasing costs and increasing capabilities of sensor equipment.

Next-generation sensors are generating structured and unstructured data much faster than OT teams can manually analyze. This situation necessitates new approaches to data management and processing.

OT teams have begun to understand the value of data. While for a long time they struggled with managing what they thought was too much data, AI tools have shown ways to transform this abundance of data into meaningful and actionable information.

Today, many AI software applications work effectively with large datasets. However, in fields like OT, which rely on continuous uptime, risk reduction, deterministic control loops, and fundamental engineering principles, the use of AI presents different challenges.

### AI Challenges in Operational Technologies

The rise of generative AI has demonstrated how successful large models can be at synthesizing and summarizing information. However, without being bound by physical realities such as thermodynamics, equipment design limits, mass and energy balances, control loop parameters, and safety interlocks, AI can make recommendations that conflict with process reality.

Such errors can erode operator trust, lead to delayed decisions, and increase risk in real-time and critical OT environments.

### AI-Powered Visibility and Predictive Capability

In increasingly complex market dynamics, OT teams expect insights and predictive analytics from AI that go beyond their current capabilities. This increases innovation opportunities while boosting operational efficiency.

For AI applications to provide accurate and useful results, it is necessary to combine computational facts with AI's high-level intelligence. This ensures effective communication between operators and production systems via AI.

### The Importance of Industrial AI

Unconstrained AI does not yield accurate results. For example, free and open-ended questions, such as a new operator simply asking about the previous shift's status during a shift change, can lead to meaningless or irrelevant data output.

Noisy and irrelevant data reduces operator trust in AI. Therefore, industrial AI must offer narrow-scope and contextual recommendations, supported by domain expertise and physical rules.

### Features of Industrial AI

• Principles-based: Relies on immutable fundamental principles such as physics, chemistry, equipment design, and control limits.
• Persona-centric: Offers specific recommendations for different user roles such as operators, maintenance engineers, and process engineers.
• Operationally contextual: Adapted to the factory, unit, equipment, and control loop hierarchy.
• Contextual data usage: Can work with raw and unstructured data, but its performance improves with meaningful data.

### Data Infrastructure for Industrial AI

A robust data infrastructure is essential. To this end, many organizations are establishing an enterprise operations platform based on a data fabric that provides seamless and contextual data movement.

This platform fluidly moves data from the field to the virtual and to the cloud, supporting the effective operation of industrial AI.

### Conclusion

The most advanced automation providers are integrating AI tools with principles and industry knowledge to make operational AI more reliable and scalable. Building OT modernization on data integration and a comprehensive data fabric opens the door to a new industrial era that will provide a competitive advantage.

This development will be one of the key elements of successful companies in the coming years.