"The Cause of the Defect Is Bolt Detachment"... Trusting Data Over Skilled Intuition Reduces Errors and Boosts Efficiency [AI Autonomous Manufacturing: Opening the Future]

"All In"? Large Corporations and Partners Join Forces for AI Transformation
"An Era Where Data Must Be Trusted Over Skilled Intuition"

"The cause of the defect is that the bolt was detached due to over-tightening by the operator in charge of harness holder assembly at Process 16. In the past week alone, about 10 cases of missed tightening defects have occurred in this process."

Inside the factory where the core component of the washing machine, the drum, is assembled, a conversation can be heard discussing the cause of a defect and how to resolve it. This voice, which explains everything from identifying the cause of the defect to the most recent defect history and the solution, is, surprisingly, AI. The operator wears a wearable device that looks like goggles and solves problems by conversing with the AI. This is a scene expected to become common on the manufacturing floor as early as three years from now.

Exterior view of Shinseong Delta Tech, a specialized company in OEM and ODM for home appliances located in Changwon, Gyeongnam. Photo by Jang Heejun

During a recent visit to the home appliance parts factory of Shinseong Delta Tech in Changwon, Gyeongnam, such AI pilot projects were in full swing. In some processes, "manufacturing AI" has already begun to be implemented, and the company is currently at the stage of collecting vast amounts of data generated on site. The goal is to build a large language model (LLM) similar to "Manufacturing ChatGPT" and, furthermore, to implement a large action model (LAM) that can understand and act on the intentions of workers.

Shinseong Delta Tech, a partner of LG Electronics, is a representative site demonstrating that AI is not a technology limited to large corporations, but is becoming a reality for "data-driven quality innovation" across the entire manufacturing supply chain.

"All In"?Large Corporations and Partners Join Forces for AI Transformation

The drum washer's tub assembly is lined up fully assembled on the washing machine production line at Shinsung Delta Tech. Previously, a person had to inspect the presence of 50 different parts one by one, but now artificial intelligence (AI) technology can even handle noise measurement. Photo by Jang Heejun

AI, now embedded in the manufacturing field, is evolving beyond simply making work more convenient to compensating for human shortcomings. Lee Donghan, CEO of Shinseong Delta Tech, explained, "The repair process is second in importance only to the team leader among operators, and it requires a high level of understanding not only of the process but also of functional issues. Through AI, manufacturing companies can accelerate the adaptation of new workers, achieve efficient workforce redeployment, and reduce costs."

A significant portion of Shinseong Delta Tech's production line has already adopted AI technology. Upon entering the factory, the drum washer's tub assemblies are lined up, fully assembled. Previously, a person had to check for the presence of 50 different parts one by one, but from July, through collaboration with LG Electronics, AI technology will be able to check the fastening status of parts and even handle noise measurement.

A robot that automatically assembles parts is operating on the washing machine production line at Shinsung Delta Tech. The vast amount of data generated by each robot and equipment on site is being collected to build a large language model (LLM) specialized for the manufacturing process. Photo by Heejoon Jang

The dryers exported to the United States have already reached an automation level of 85%. Human workers are only responsible for feeding parts and receiving the final assembled products. Production output reaches 600 units per hour. Data from robots and equipment installed at each process is collected in real time to build LLM and LAM systems. The amount of data gathered on site has increased significantly, from 3GB to 12GB per day. Jeong Jinwoo, Executive Director at Shinseong Delta Tech, explained, "By introducing robots, we have automated 45% of the production process and digitized 30% of tasks such as product information management and production scheduling. This has not only improved productivity but also reduced labor dependency by about 16%."

As Chinese companies rapidly gain competitiveness, Korean manufacturing firms are facing worsening profitability. In an era where "manufacturing capability equals corporate competitiveness," failing to maintain productivity threatens survival. Shinseong Delta Tech sees rapid AI transformation (AX) as a breakthrough to overcome the manufacturing crisis.

Robots and workers are jointly operating on the washing machine production line at Shinsung Delta Tech. The vast amount of data generated by each robot and equipment on site is being collected to build a large language model (LLM) specialized for the manufacturing process. Photo by Jang Heejun

Executive Director Jeong Jinwoo said, "I tell our team members to imagine a daily routine where, after clocking in and enjoying a cup of coffee, they simply hit 'Enter' on the master system and instantly see the previous day's production report. We can no longer afford to have people check data and make phone calls just to find out what happened last week or last month. We need to make fast and accurate decisions based on technically refined data."

CEO Lee repeatedly emphasized the importance of win-win cooperation with large corporations. He said, "In reality, companies can only make new investments if the return on investment (ROI) is realized within two years, but for difficult processes where AI technology is to be applied, it takes about four to five years for investment effects to appear. In addition to government support, AI technology must spread throughout the supply chain through inter-company cooperation to secure industrial competitiveness."

"An Era Where Data Must Be Trusted Over Skilled Intuition"

A view of the small and medium-sized enterprise Gwangwoo factory located in the middle of the steel industrial complex in Pohang, Gyeongbuk. Provided by Gwangwoo

The production plant of Kwangwoo, a small and medium-sized enterprise that produces the essential special lubricant "synthetic ester" for steel rolling equipment, is located in the steel industrial complex in Pohang, Gyeongbuk. Instead of the characteristic pungent smell of organic solvents, the space was filled with the quiet sound of machines. Kwangwoo, a partner of POSCO, Korea's largest steelmaker, has implemented a system where AI, instead of people, precisely controls the amount of raw materials added.

The ester process manager guided the reporter through the synthetic process building, explaining, "The amount of raw materials fed into the machines is automatically controlled within a margin of error of 0.1%. In the past, people had to manually adjust the valves, which led to high fatigue and frequent mistakes."

Synthetic process line of Gwangwoo. Major variables such as temperature, flow rate, and pressure are automatically controlled in the large mixer and purifier. Photo by Seongpil Cho

Kwangwoo began process automation and AI adoption in 2015. Currently, the company is in the midst of a second phase of advancement, where key process variables like temperature, pressure, and recovery amount are controlled in real time. All data is collected in real time and consolidated into a digital dashboard, which is also used to detect early signs of equipment abnormalities.

The core process is "blending quality control." If key figures such as raw material mixing ratios or blending temperatures deviate from the standard, the AI-based system immediately issues a warning. The standard values are set by analyzing past production history based on POSCO's required quality conditions. Park Taejoon, CEO of Kwangwoo, said, "In the past, improvements were only made when customers raised issues, but now we identify deviations internally first."

BT-12 process dedicated operation panel screen. Equipment operation information and warning system are visualized in real time. Photo by Seongpil Jo

The blended base liquid is sent to a purifier, where impurities, color, and oil separation are checked again. If an error occurs, the system automatically sends an alert. In the final packaging stage, manual work and automatic inspections are performed in parallel. Kwangwoo produces dozens of products in small, customized batches. For this reason, quality control at the individual lot level is more important than repetitive automation. CEO Park said, "Special lubricants are highly sensitive to quality. Now is an era where we must trust data over the intuition of skilled workers."

Since the introduction of AI quality management, the rate of customer complaints has been reduced to less than half, and quality variation due to operator skill level has also decreased significantly. The factory at Kwangwoo does not resemble the busy, fully automated scenes with robots. Instead, the collected data is the core asset. The focus is on an information system that analyzes and manages quality, equipment, and history data in real time. A field employee said, "Right now, we are at the 'civil engineering' stage for AI and machine learning. Machines can only function properly once the foundational data is in place."

A Kwangwoo production team member explaining the BT-12 blending process. Each tank operates based on blending data for each product. Photo by Seongpil Cho