6894d0777a513573f997fae3 Adobestock 1258426465

The role of AI in industrial data management

Aug. 7, 2025
Artificial intelligence helps industry on its path toward autonomous operations

The industrial sector is steadily progressing toward autonomous operations by enhancing existing automation technologies with artificial intelligence (AI). Every organization is working toward this goal, and technology providers across various domains contribute to the transformation. Data plays a dual role in the journey—it’s both the fuel and the foundational element. However, we must first understand and manage its context to harness its full potential.

Fragmented data challenge

One of the major hurdles faced by industry is fragmented data. Industrial environments rely on a wide array of systems—DCS, PLC, HMI, SCADA and IoT platforms—that support real-time monitoring, control, trending, etc. These systems manage machine logic, process control and work order execution. Additionally, manufacturing systems manage operations, resource planning, workflow coordination, order tracking, quality reporting, root cause analysis and yield accounting.

Beyond operations, innovation systems analyze historical data, model and optimize processes, support R&D and plan services. Regulatory systems manage risk modeling, compliance reporting and strategic planning. Enterprise systems focus on business, customer, product, asset and process orchestration. The coexistence of these diverse systems leads to data silos, making integration and utilization difficult.

Diverse data types and underutilization

Industrial operations generate several types of data:

  • Structured: reports, transactions, work orders;
  • Semi-structured: design and engineering files, digital simulations (2D/3D);
  • Unstructured: video and audio recordings;
  • Event: typically structured; and
  • Geolocation: often semi-structured.

Despite the abundance of data—from sensors to control systems to historians—it remains underutilized. For instance, while sensor values with process variables and/or manipulated variables (PV/MV) are sent to a DCS and diagnostic data goes to asset management systems, sensor performance history may only be accessible to manufacturers. End users often rely on local historian systems, missing valuable insights for predictive analysis.

Collaboration and integration barriers

The diversity of data sources and missions drains resources and complicates collaboration. Data silos restrict access, reduce interoperability and compromise data integrity. Ensuring data is accessible and interoperable across applications, systems, users and suppliers is a significant challenge.

Get your subscription to Control's tri-weekly newsletter.

Cybersecurity adds another layer of complexity. Protecting data at its source is a crucial step to prevent manipulation and mitigate threats.

Flattening the Purdue Model

Advances in cloud computing, digital twins and AI enable better data utilization, operational efficiency and decision-making. A key trend is flattening the Purdue Model—integrating levels 2, 3 and sometimes level 4—to break down silos and enhance data management.

Organizations such as CESMII (Smart Manufacturing Institute), in collaboration with OPC Foundation, VDMA and the Industrie 4.0 Platform are driving standardization and interoperability to build a collaborative ecosystem.

Read more about the Purdue Model:

Data contextualization

Effective data management is foundational to autonomous operations. It involves:

  • Data cleaning: removing noise and inconsistencies;
  • Data classification: organizing data by type and relevance; and
  • Data contextualization: understanding relationships and meaning.

Contextualization, often referred to as data fusion, is best achieved through knowledge graphs, which visually represent data relationships. They can be large and complex, or modular with orchestration layers.

Once contextualized, data can support applications that provide recommendations based on operational needs, asset conditions, compliance requirements and environmental factors. Clear objectives and guardrails are essential when applying AI technologies.

OT and IT collaboration

Achieving autonomous operations is not a solo effort—it requires collaboration between operational technology (OT) and information technology (IT). OT brings domain expertise, while IT contributes AI methodologies. Building an open ecosystem that leverages the strengths of all players is key. The partnership between Yokogawa and KBC exemplifies this synergy, combining deep operational knowledge with advanced analytics.

AI Agents and decision support

Beyond data management, the next step is leveraging AI to support decision-making, which involves building AI agents using generative AI and agentic AI frameworks. These agents function as building blocks for applications, orchestrating multiple agents to deliver intelligent, autonomous solutions.

Building the digital fabric

The path to autonomous operations involves creating a digital fabric that connects and contextualizes industrial data using AI. By breaking down silos, managing data effectively and fostering collaboration, industry can enhance operational efficiency and empower smarter decision-making.

About the Author

Penny Chen | Yokogawa

Penny Chen is a senior technology strategist at Yokogawa and a Control 2020 Process Automation Hall of Fame honoree.

Sponsored Recommendations

Municipalities are utilizing inline total solids measurements to enhance sludge thickening, lower polymer usage and cut operational expenses.
Carbon dioxide is increasingly recognized as a vital resource with significant economic potential. While the conversion of carbon dioxide into products is still in its infancy...
Discover our wide range of temperature transmitters that convert sensor signals from RTDs and thermocouples into stable and standardized output signals!
An innovative amine absorption-based carbon capture process enables retrofitting of existing industrial facilities to reduce emissions in hard-to-abate sectors, with advanced ...