With People
AI-Driven Innovation in Manufacturing Environments
AI technology is now integrated into all areas, from industries to everyday life. It has moved beyond the realm of science fiction and become a reality. With AI-related technologies recently earning Nobel Prizes in Physics and Chemistry, their impact on the global scientific community is profound. In the future, a company’s competitiveness will depend on how quickly and strategically it can apply AI technologies in the right areas.Recognizing the potential of AI early on, LG Innotek began research on applying AI to its manufacturing process in 2019. Seunghun Yu, Team Lead of the AI/Big Data Solutions Team, explains how Industrial AI drives innovation in manufacturing and shares his journey in developing Industrial AI. Q. Please briefly explain what Industrial AI is.Industrial AI, also known as ‘Manufacturing AI,’ uses artificial intelligence to replicate the physical environment of manufacturing processes. It enables predicting product quality, managing maintenance, and optimizing manufacturing processes. To explain in detail, AI analyzes the entire manufacturing process to detect potential defect points in advance. Then, it runs simulations based on various process variables to determine the optimal manufacturing conditions that minimize defect rates and significantly enhance manufacturing efficiency.Q. What was your role in the development of Industrial AI?The Optical Solutions Division has long recognized the importance of data-driven process optimization to improve product quality and has developed relevant technologies. In 2019, we identified the need for AI-powered quality prediction and initiated advanced research into related AI technologies. By 2021, I took on the role of leading the newly established R&D team, carrying out projects focused on developing this technology.At the time, Industrial AI was still in its early stages, with limited resources and minimal research on its concepts and methodologies, both within the company and in academia. In this situation, my role was to develop LG Innotek’s unique concepts and technologies for Industrial AI to overcome these technical barriers. Q. Since AI is still in its early stages, developing Industrial AI must have come with challenges. Could you share the challenges you encountered and how you overcame them?Unlike general AI, Industrial AI is a specialized technology developed specifically for the manufacturing sector. It differs from ‘Neuro AI,’ commonly studied in academia, which mimics the human brain. Instead, Industrial AI replicates physical environments. While Neuro AI primarily processes image and text data, Industrial AI not only handles these but also incorporates a variety of physical factors into its data.Given the specialized nature of this technology, there were very few relevant references available during its development. There were no research papers specifically addressing technologies for the manufacturing sector. It felt like venturing into a new field, encountering numerous inevitable difficulties and failures along the way. As I mentioned earlier, academic data structures are quite different from those used in real-world manufacturing environments, making it challenging to integrate AI models effectively with a deep understanding of our data. Even when we applied academic findings, the performance of the AI didn’t meet our expectations, and many issues remained unresolved despite thoroughly reviewing recent research papers. Putting much effort into trying various approaches and not seeing much improvement was mentally exhausting for both me and my entire team.Despite these challenges, we stayed motivated by reminding ourselves that we were pioneering new technology, believing that overcoming these obstacles would position us as industry leaders. Through continued experimentation and refinement of AI models and data, we finally started achieving the results we had worked so hard for. Although I don't usually have emotional swings, every time we overcame a technical challenge together, I found myself overwhelmed with joy. Q. Since Industrial AI development involves collaboration of hardware and software expertise, could you describe how experts from different fields work together on this project?Developing Industrial AI required collaboration across various teams. The AI team in the CTO division had strong technical expertise in AI, but they lacked experience with real-world manufacturing data. Meanwhile, experts in optics, substrates, manufacturing, and development had deep knowledge in the data and their fields but were less familiar with AI technologies.To close this gap, we formed cross-functional teams from the beginning. Through the TDR (Tear Down & Redesign) framework, members from the CTO and business division worked together, sharing knowledge and insights for the project. Instead of traditional project management methods, we adopted Agile methodologies like Scrum. This project management approach divided the project into manageable tasks and smaller achievable milestones with short sprint cycles. Each 2-week sprint included planning, development, review, and retrospection, enabling us to progress steadily. This process also involved executives from each division, not just the engineers, who participated in task reviews and adjusted the team’s direction as needed to keep us aligned with our goals. Through this, we could collaborate smoothly and effectively.Q. What customer pain points could LG Innotek’s Industrial AI resolve?From a quantitative perspective, our Industrial AI reduced F-Cost(Failure costs, expenses incurred due to defective products in manufacturing) through defect minimization, while from a qualitative standpoint, it improved product quality, resulting in greater customer satisfaction. Additionally, Industrial AI represents a technological innovation that our competitors have yet to attempt, highlighting LG Innotek’s unparalleled technical expertise and capabilities. I believe this has strengthened customer trust and solidified their decision to choose LG Innotek.Furthermore, Industrial AI is a scalable technology. Its applications are not confined to a single business unit but extend across multiple areas, such as substrates and automotive electronics. I see it as a foundational technology that can provide greater value to our wide range of customers. Q. Lastly, why do you think Industrial AI is remarkable?Today, most AI research focuses on LLMs(Large Language Models) and generative AI, while Industrial AI research in the manufacturing sector remains relatively limited. In particular, it’s almost impossible to conduct research in academia due to restricted access to essential manufacturing data. In this context, LG Innotek’s R&D investments and active application of Industrial AI make our technology at the forefront of this specialized field.I believe that innovation happens when we push beyond the boundaries. Both our business division and CTO collaborated tirelessly together for a long time, overcoming numerous challenges and failures. Now, we’re starting to see the fruit of those efforts. I think LG Innotek’s culture of always encouraging attempts, embracing failure, and striving for continuous improvement has been the powerful driving force behind these achievements. Although Industrial AI is still in its early stages, I’m confident that it holds enormous potential for creating more value in the future. All the journey we've been through and the infinite possibilities ahead of us make Industrial AI truly remarkable.