Smart Factory Provides Data to Improve Productivity
Without the use of AI technology, it would be nothing more than simple automation.
Need for AI solutions and talents that respond to specific cases


Manufacturing is the core of the domestic economy. Korea is one of the world's top five manufacturing countries along with the United States, China, Germany, and Japan, and among them, manufacturing accounts for the largest proportion of the entire industry, at 30.4% as of 2017. This is why institutions and companies are focusing their attention on smart factories.

The successful distribution of smart factories can serve as a stepping stone for improving domestic manufacturing capacity and national competitiveness. The government defines smart factories as intelligent factories that apply information and communication technologies (ICTs) such as IoT, AI, and big data to all or part of the manufacturing process from planning and design to production, distribution, and sales to improve corporate productivity and product quality.

Automation in manufacturing is still ongoing. Automation is minimizing human intervention in the manufacturing process and increasing the role of machines. As a result, it aims for an unmanned factory. It is often said that a smart factory is different from automation. It is an ICT-based factory that plans and produces customized products for customers. Is this feasible?

Since purchasing decision makers are humans, product planning and design are uniquely human domains. Smart factories can only play a more limited role than the definition explained above, processing and providing data accumulated in the manufacturing, logistics, and sales processes so that humans can create customized products for customers. Even this is a step further than automation and is difficult to achieve.

◇ AI technology must be applied to be called a smart factory

Manufacturing process data is collected by sensors. There are various types of sensors on the market, but many capture physical conditions such as light, force, sound, electricity, temperature, and humidity. They are excellent for collecting operational data and detecting abnormal signs of factory equipment connected to the Internet, but to obtain product-related data, an AI system that recognizes the state of the object is required.

AI systems combined with image sensors and other technologies verify the quality of new products being produced by comparing them with data from good products. In addition, AI technology can be utilized throughout the entire manufacturing process.

The Ministry of Trade, Industry and Energy divided the smart factory level into four stages: basic, intermediate 1, intermediate 2, and advanced. The basic smart factory can digitize production information and manage product production history, but AI technology is not utilized. AI technology is essential for intermediate 1, which can collect and analyze production information in real time, intermediate 2, which can control the production process through the system, and advanced, which can integrate the entire manufacturing process and produce customized products.

◇ Smart Factory = Field Expertise + AI Expertise

The Ministry of SMEs and Startups and nine other government ministries announced that they would supply 30,000 smart factories to SMEs and increase the operating workforce to 100,000 by 2022. According to the National IT Industry Promotion Agency (NIPA), as of 2019, there were no factories that had reached the advanced stage, and the proportion of factories at the basic stage reached 76.4%. This is limited to simple automation. As of January 2021, the number of smart factories nationwide under the popularization project is 19,799.

The situation is no different for large corporations. Since 2019, McKinsey and the World Economic Forum (WEF) have been selecting manufacturing plants utilizing the 4th Industrial Revolution technology as “lighthouse factories.” Currently, there are 69 lighthouse factories around the world, and only one in Korea is POSCO. The Ministry of SMEs and Startups benchmarked this and started the “K-Smart Lighthouse Factory” project, selecting 10, but unlike the WEF, which evaluates performance, it evaluated plans and was criticized for being a duplicate support project.

The reason why it is difficult to introduce smart factories is because there is a lack of AI solutions and experts.

In April, the Korea Institute for Industrial Economics and Trade announced a survey on the status of AI utilization in 283 companies. Of these, 53% cited “difficulty in hiring personnel with the right skills” as a barrier to AI adoption and utilization. 32.2% cited funding issues, 25.1% cited infrastructure issues, and 16.6% cited training issues for existing employees. According to Statistics Korea, only 409 out of 13,255 domestic companies had adopted AI technology into their businesses as of 2019.

Even if AI solutions and experts are secured, it is not easy to apply them to actual fields. Since each company has different business areas and methods, it is necessary to determine where AI solutions should be used to achieve effectiveness. Even in the manufacturing industry, the industry is divided into electronics, steel, food, assembly, etc., and each operates its own process such as R&D, material management, production, sales, and service in its own way.

Although data standardization and protocol unification for smart factory linkage can be achieved with a central organization, each smart factory operation will inevitably be specialized. Specialized AI models must be applied to specialized factories, but it is difficult for field experts without AI expertise and AI experts without field expertise to create specialized AI models.

◇ Smart factory construction solutions and services in the AI industry

Lack of human resources and funds are major obstacles to AI adoption by companies, and even when both conditions are met, it is still difficult to develop AI models specialized for the company. These difficulties can be resolved to some extent by collaborating with AI technology companies or using their solutions.

The three domestic mobile carriers, SK Telecom, KT, and LG Uplus, which are exploring applications for 5G mobile communication technology, have selected smart factories as their target and are developing related businesses.

SKT's 'Metatron Grand View' is an AI, cloud-based subscription smart factory service. It supports services such as real-time monitoring, facility abnormality and failure alarms, and facility-specific AI analysis modeling after 3 to 6 months of data accumulation. Dongyang, Seongshin Compressor, etc. are using it.

KT plans to launch a new product that links its ‘Factory Makers’ platform with three collaborative robot products from Hanwha Machinery in the third quarter. In July, U+ launched a smart factory solution that diagnoses and checks for malfunctions and failures in factory equipment in advance with AI specialist OnePredict. It is also signing MOUs with logistics and city-related technology companies.

The three mobile carriers are securing as many solutions as possible in the smart factory sector to maintain influence and increase market share in the future market expansion phase.

MathWorks strengthens the AI capabilities of engineering software such as MATLAB and Simulink with each new release, enabling even non-AI experts to create and train AI models. It also supports code generation and distribution in C, C++, HDL, PLC, CUDA, Java, and Python for models developed according to cases.

Schneider Electric released the ‘EcoStructure Machine’ platform and ‘EcoStructure Machine Advisor’ solution to the domestic market in July. This 'MS Azure' cloud-based facility digitalization solution provides remote equipment management, tracking, and data analysis functions to make factory operations more intelligent. When selling its dust collectors, domestic company NIT Korea provides customers with a management solution based on the solution it produced together.

LS Electric is providing customized services to each manufacturing company based on its 'Tech Square' smart factory platform, and has also signed an agreement to contribute 3.3 billion won to the Korea Foundation for Large and Small Businesses and Agriculture and Fisheries Cooperation to support the establishment of smart factories for large and small businesses.

Xilinx supplies FPGA products equipped with Domain Specific Architecture (DSA) so that companies that plan to develop AI-based businesses can easily apply the latest AI technology to existing equipment. It also provides AI frameworks such as ‘TensorFlow’ and ‘Pytorch’ that support this, as well as the ‘Vitis™’ integrated software platform that supports C++ and Python.

◇ AI support measures are needed for small and medium-sized enterprises that make up the majority of manufacturing companies

Manufacturing powerhouses such as Germany's 'Industry 4.0', the US's 'Advanced Manufacturing Partnership 2.0', Japan's 'Society 5.0', and China's 'Made in China 2025' policy are accelerating manufacturing innovation. The non-face-to-face, remote control trend due to the pandemic also accelerated this.

We must not fall behind either, but the current reality of the Korean manufacturing industry is that out of 69 lighthouse factories around the world, only one is located in Korea.

Manufacturing is moving from automation to autonomy, and AI technology is at the core of it. It is urgent to develop AI talent training plans, jobs for them, and AI products, solutions, and services that can be easily used by small and medium-sized enterprises that make up the majority of the manufacturing industry.