Kia (000270.KS) has reduced annual labor hours for equipment maintenance by 3,750 hours since introducing digital twin technology at its electric vehicle production plant in August 2024, according to an analysis. The operational innovation was achieved through real-time monitoring of equipment status and anomalies in a computer-generated virtual space.
According to Hyundai Motor (005380.KS) Group data obtained by The Seoul Economic Daily on January 12, the man-hours required for body production line maintenance at Kia's Gwangmyeong EVO Plant in Gyeonggi Province dropped to one-third of pre-digital twin levels. Man-hours refer to work volume converted into total labor time, a key metric used in manufacturing to measure labor input and labor costs.
Previously, addressing equipment malfunctions on body production lines required an average of four workers and 350 man-minutes. At the Gwangmyeong EVO Plant, resolving a single error now requires four workers and only 100 man-minutes. Based on these calculations, Kia concluded that 3,750 hours of man-hours were saved annually in addressing equipment malfunctions. The figure assumes 900 annual equipment malfunctions, multiplied by the 250 minutes saved per incident. For Kia, this means freeing up 3,750 hours of labor annually for redeployment to other tasks.
The efficiency gains at the Gwangmyeong EVO Plant represent a prime example of digital twin applications in manufacturing. Digital twin technology replicates real-world objects and spaces as data in a computer environment, creating a virtual "twin." Kia established a system to monitor equipment malfunctions and anomalies in real time through virtual space, streamlining maintenance operations. Hyundai AutoEver (307950.KS), the system integration affiliate of Hyundai Motor Group, built the digital twin system for the Gwangmyeong EVO Plant.
Beyond Kia, major Korean manufacturers in semiconductors and steel are adopting digital twin technology, establishing it as an essential tool for manufacturing innovation alongside artificial intelligence and robotics. The manufacturing industry expects that workflow design utilizing real-time data integration and virtual predictive simulations will drive efficiency improvements across entire processes.
The two pillars enabling operational efficiency are real-time data integration and simulation. Real-time data linkage between actual equipment and virtual space eliminates spatial constraints for workers, enhancing the quality of remote collaboration in facility management and control. Furthermore, simulation serves as a core function for proactive efficiency improvements. Companies can predict outcomes of environmental changes by modifying equipment and facility layouts in virtual space. Simulation allows companies to examine multiple scenarios within virtual environments and devise optimized workflow designs.
The integrated digital twin control system at Kia's Gwangmyeong EVO Plant focuses on real-time data integration. Kia and Hyundai AutoEver implemented "black box" and WebRTC (real-time communication) technologies at the plant. The black box is a digital recording system. When anomalies occur in the factory environment, real-world data is reported to the virtual computer environment in real time. Workers can review black box records to identify the exact timing of problems and faulty components. Similar to reviewing dashcam footage to determine the cause of a traffic accident, workers can examine digital twin records as video to diagnose malfunction causes.
WebRTC enables viewing digital twin control systems, including the black box, through web environments. Typically, using industrial digital twin programs required computers directly connected to high-specification graphics processing unit (GPU) servers. WebRTC implements digital twin programs in web environments, creating a multi-access system accessible through smartphones, tablets, and various devices. This enables workers in different locations to share the same virtual space data and collaborate.
The black box and WebRTC had an immediate impact on operational efficiency by reducing time required for malfunction reporting, anomaly identification, and troubleshooting. The number of personnel and time required for initial response after malfunctions decreased significantly. According to Hyundai Motor Group data, anomaly identification previously required four workers for approximately 30 minutes, equaling 120 man-minutes. At the Kia EVO Plant, one worker can complete anomaly identification in 10 minutes—a reduction from 120 to 10 man-minutes.
Beyond Kia, other companies are actively utilizing simulation capabilities. Hyundai Steel (004020.KS) is working with Hyundai AutoEver to build a digital twin control system for responding to gas leak accidents at cold-rolled steel plants. The company is developing a platform that manages workers, equipment, and workplaces in a single virtual space platform to enhance safety. The platform will analyze data from gas detection sensors in real time within virtual space and track the origin of hazardous gas leaks. A gas dispersion path simulation function is also under development to support decision-making for preventing secondary accidents.
Additionally, semiconductor manufacturers Samsung Electronics (005930.KS) and SK hynix (000660.KS) are partnering with Nvidia to pursue digital twin innovation. Samsung Electronics uses Nvidia's digital twin platform "Omniverse" to detect real-time anomalies in semiconductor production equipment and optimize production schedules in virtual space. SK hynix is also building virtual semiconductor factories based on Omniverse to simulate production flows and material movement.
The digital twin adoption trend is expected to intensify competition among companies for data acquisition. Digital twin experts cite ground truth data and reliable golden data sets as prerequisites for manufacturing innovation. Attaching small computers to facilities to collect process data and extracting high-quality data has become increasingly important.
"The ultimate reason for using digital twins is to find accurate answers through simulation predictions," said the CEO of a digital twin platform developer. "Obtaining the most refined data from the field is essential to achieving simulation results close to the correct answer."
