Hyundai and Kia connect metrology to software-defined factories

By Sophie Lindqvist

At the seventh E-FOREST TECH DAY 2026, Hyundai Motor and Kia showcased about 200 manufacturing technologies at the Uiwang Research Institute in South Korea.

The programme covered software-defined factories, automation, energy management and future mobility. Its stated aim is to connect data, AI and standardised software through common infrastructure, extending beyond fixed automation into managed, optimised production.

For buyers, the immediate finding is a missing performance baseline. The supplied announcement gives no inspection accuracy, defect-detection rate, false-reject rate, resolution, sampling rate, calibration interval, cycle time or measured productivity gain. Claims about better quality and easier scaling therefore remain intentions, not measured results.

Takeaway: E-FOREST shows the intended architecture, but not the acceptance data needed to judge it.

What does the common digital foundation change?

Hyundai Motor and Kia describe E-FOREST as a smart-factory initiative that links production processes and technologies. Its latest work establishes a unified framework for collecting and managing data, plus a common development environment, to simplify application development, deployment and scaling across plants in Korea and international markets.

For multi-site manufacturers, shared data structures could reduce fragmentation among equipment, inspection systems and production-management software. A common framework could let teams transfer proven applications, adapt them locally and respond faster to changing requirements, making production capabilities more configurable.

Digital twins form another layer. By reproducing factory equipment and processes virtually, the twins let engineers simulate line configurations, find process constraints and assess proposed changes before touching live production. Reliable feedback from physical equipment then allows expected performance to be compared with actual manufacturing results.

The deciding principle is model fidelity: a digital twin is useful only when its virtual state stays tied to traceable physical measurements. For dimensional metrology (measurement of part geometry), inspection results, equipment status and process context can show whether a component meets specification and which conditions affect it. The announcement names no interface standard or schema revision, so interoperability remains an implementation requirement.

Takeaway: Common data and faithful digital twins can shorten deployment, provided each plant supplies consistent, traceable measurements.

What will the AI agents actually do?

Hyundai Motor and Kia are developing factory-specific AI agents for three operational tasks:

• retrieve and interpret manufacturing data;
• search accumulated manufacturing know-how;
• create equipment manuals.

This moves AI beyond isolated analysis toward everyday manufacturing support. The agents could give production staff faster access to instructions during faults and investigations. With verified production data, they could support troubleshooting and expose relationships among equipment behaviour, process conditions and product quality.

That promise depends on context, not database size. Data integrity and traceability must separate verified information from uncertain conclusions. In quality-critical work, every recommendation needs a defined validation and decision process before an operator or controller acts.

Conference and partner sessions focused on practical applications and future strategies, covering physical AI, data-driven manufacturing innovation and manufacturing AI. Physical AI links AI with systems such as robotics, automated handling and adaptive production that interact with the physical world; reliable perception, control and safe responses determine its industrial use.

For metrology, measurement technologies must resolve relevant geometry or quality characteristics and deliver results confident enough to trigger process actions. That closes the loop from data collection to control, rather than leaving instruments as report generators.

Takeaway: The AI layer can shorten knowledge retrieval, but validation must travel with every recommendation.

Can vehicle inspection become a process signal?

The programme featured a tunnel-type vision system for automated vehicle exterior inspection. Hyundai Motor and Kia developed it as an alternative to third-party systems, using 3D simulation. Plans call for deployment at plants in Korea and overseas, with the stated intent of improving finished-vehicle quality.

That is a roadmap, not a measured result. The supplied material specifies no defect threshold, detection rate, false-reject rate, imaging configuration, cycle time or calibration interval. Buyers should treat 3D simulation as a method label, not evidence of production-speed detection for a defined defect class.

The larger opportunity is integration. Optical inspection can examine surfaces for potential defects at production speeds that would be difficult to achieve manually. Joining results to equipment status and process records lets inspection support monitoring, defect analysis and continuous improvement instead of remaining a final pass/fail checkpoint.

Standardised interfaces could let plants compare quality performance across lines and facilities, identify recurring problems and investigate their causes. The same requirement applies to coordinate measuring machines (CMMs), optical measurement technologies and in-process inspection: consistent formats and reliable results are prerequisites for automated quality decisions.

Takeaway: The inspection system matters only when its output is fast, quantified and linked to process conditions.

Where do energy and new manufacturing processes fit?

Beyond the core digital work, E-FOREST presented assembly and inspection automation, energy solutions and technologies for future mobility. Connected energy monitoring and data analysis can expose consumption inefficiencies and support sustainable production by linking resource use to factory context.

The exhibition also included new forming methods and additive manufacturing, which builds components directly from design data without conventional moulds. These routes can increase manufacturing flexibility where component complexity, tooling requirements or production volume influence process selection.

Inspection and verification remain necessary because new processes can introduce dimensional variation and material inconsistency. Measurement results must feed back into production so engineers can adjust the process, not merely document a nonconforming part.

Takeaway: Software-defined production still depends on physical verification.

What must be proven before adoption?

The transition succeeds only if Hyundai and Kia integrate new tools with installed equipment, enforce consistent data standards and demonstrate measurable gains in quality, productivity and operational flexibility. Metrology suppliers must therefore act as connected infrastructure, not isolated report generators.

Takeaway: Before adoption, can every measurement and AI recommendation be traced to verified source data, validated against production outcomes and governed by a defined decision process?