How Data Elements Drive Continuous Growth in Manufacturing: Challenges and Solutions

Background

Data, originally a symbolic representation of human practice, has evolved into a business‑colored numerical asset that enables new methods for understanding complex physical and operational systems. In the era of digital and intelligent transformation, data continuously loops through production cycles, driving richer, more diverse business scenarios.

Three Data‑Driven Challenges

The rapid increase in data scale, variety, and complexity creates three core challenges:

Explosion of application scenarios : Traditional industrial information systems listed a limited set of production‑management problems, but digital transformation generates countless new data‑driven use cases.

Enrichment of business‑application scenarios : As data feeds more processes, aligning data with rapidly evolving business needs demands agile, high‑speed development beyond conventional software cycles.

Expansion of intelligent‑application scenarios : Leveraging AI large models and domain knowledge to uncover hidden value requires new methods for data integration and model deployment.

Exploration and Practice in the Full Manufacturing Process

We moved from the traditional ISA‑95 multi‑layer architecture to a flat, industrial‑internet‑based platform. The platform unifies data collection, aggregation, and governance, turning raw data into reusable data assets. These assets support a growing ecosystem of business and intelligent applications, from mobile factory apps to PC dashboards.

Key steps include:

Standardized design and modeling to create common data standards, naming conventions, and metric systems.

Construction of a unified data service catalog that makes assets discoverable, understandable, usable, and controllable.

Integration of IT (enterprise) and OT (operational) data, handling structured enterprise data and multimodal sensor data.

Enriching Business Application Scenarios

By providing generic integration components, industrial reporting tools, and mobile factory applications, the platform enables rapid creation of over 4,000 data‑driven or intelligent services. Agile tooling allows developers to iterate quickly, meeting the fast‑paced demand for new scenarios without the long lead times of traditional software projects.

Intelligent Application Scenarios with AI Large Models

We built an industry‑vertical visual AI model for steel manufacturing. Examples include:

Furnace flame recognition : Machine‑vision cameras detect flame states (e.g., arcing, spattering) and feed real‑time alerts to control systems, improving safety and efficiency across more than ten steel plants.

Dust‑control optimization : Vision‑based dust detection dynamically adjusts fan power, reducing energy consumption while meeting environmental standards.

To overcome low efficiency of single‑scenario models, we pre‑trained a large visual model on extensive steel‑industry data, then applied transfer learning to create specialist “expert” models. Knowledge distillation produces lightweight edge models deployable on smart cameras, enabling real‑time inference at the plant floor.

Conclusion

Data‑driven manufacturing demands unified governance, flexible platform architecture, and AI‑enhanced intelligent applications. By treating data as a core production factor, enterprises can break data silos, accelerate scenario development, and achieve higher value from digital transformation.