Advantech’s Linkou AIoT Co-Creation Campus Drives Smart Innovation: Four Generative AI Applications Accelerating Transformation

How to apply generative AI is a question being asked by many companies over the past two years. Advantech, a leader in driving smart transformation, has found its answer at the Linkou AIoT Co-Creation Campus. By integrating WISE-IoT iFactory software and hardware solutions with generative artificial intelligence (GAI) technology, the Linkou AIoT Co-Creation Campus has developed four major GAI applications: AI Assembly Bottleneck, OEE Assistant, AI Maintenance Assistant, and Material Management Agent. These applications effectively enhance production line efficiency, boost employee productivity, and improve supply chain responsiveness.

Utilizing Data as the Basis to Develop Various GAI Applications

Industrial PC manufacturing is characterized by small-batch, highly diverse production. Most clients place recurring orders on fixed cycles, which often means production resumes months after the initial run. This irregular cycle makes it challenging to apply data effectively. Notes Song, Manager at Advantech, noted that traditional methods like deep learning or machine learning were previously used to build AI models for data analysis, but their effectiveness was limited because they did not reflect the real production state of an assembly line. Today, GAI leverages extensive language models supported by vast data repositories. By integrating plant data with existing workflows and fine-tuning these large models, even irregularly collected data can be effectively utilized.

Building on this foundation, the Linkou AIoT Co-Creation Campus began developing various GAI applications by first identifying the needs for production management data needs and the existing pain points. GAI was then strategically applied to address these issues, yielding the following four key applications.

AI Assembly Bottleneck Agent and OEE Agent: Achieving Double-Digit Production Efficiency Gains

The AI Assembly Bottleneck and OEE Assistant applications were developed to address efficiency challenges on SMT and assembly production lines. James Chang, Project Manager at Advantech, explained that, in the past, administrators could only identify bottleneck stations by manually reviewing production line monitoring dashboards. They would then analyze the underlying data one by one to determine the root causes of anomalies and develop countermeasures. This process was not only time-consuming but also subject to variability in execution effectiveness due to differences in human judgment.

Now, with GAI analyzing production line data, early warnings are issued before bottlenecks occur, accompanied by comprehensive analytical reports. These tools assist administrators in verification, resolution, and optimization, significantly reducing the time spent on data collection, organization, and analysis. Decision-making processes have been standardized, allowing for quicker identification and resolution of production line issues. According to internal statistics from the campus, the time required to handle anomalies on SMT production lines has decreased by 26%, while assembly productivity has increased by 10%.

AI Maintenance Agent and Material Management Agnet: Enhancing Workforce Productivity 

The AI Maintenance Assistant application integrates with existing workflows and interacts with maintenance personnel through GAI, resulting in a 10% improvement in repair efficiency. With this tool, repair requests for defective products are sent directly to the AI, which evaluates the issue, provides corresponding recommendations, and generates a list of necessary materials or consumables. After the maintenance personnel verify the list, GAI seamlessly transitions to the material requisition process, reducing time spent on administrative tasks and expediting repair completion.

Additionally, GAI significantly reduces the learning curve for maintenance personnel. Chang explained that Advantech’s diverse product lineup often confronts technicians with unfamiliar issues. By providing appropriate recommendations, GAI enables even newly hired personnel to quickly adapt and perform effectively.

In Material Management Agent, GAI is applied to specific supply chain nodes, automating tasks that were previously handled manually. This eliminates the traditional step-by-step communication of needs—from demand to production to material supply—dramatically enhancing responsiveness. As a result, the system can better adapt to the rapid changes in supply chain dynamics.

Implementing AI Agent: User-Driven Development of 12 Innovative Applications

Chang emphasized that the Linkou Intelligent Campus prioritized two key strategies when integrating GAI applications: first, analyzing workflows to ensure the applications align with real-world usage scenarios; and second, establishing expert-specific databases tailored to each application type to maximize the effectiveness of the GAI solutions.

"Building on this success, Advantech is now exploring how to accelerate the development of more GAI applications," added Song. A user-centric approach has been key to driving GAI adoption, leading to the establishment of an AI Academy at the Linkou AIoT Co-Creation Campus. This initiative selects AI-enthusiastic employees from various departments as seed members. Leveraging Advantech's WISE-AI Agent solution, the team has developed 12 GAI application projects, including capacity analysis, test program generation, FQC checklist creation, production material control, material correlation analysis, PLM inquiries, and intelligent customer service.  

Ashley Peng, Product Manager at Advantech, further elaborated that WISE-AI Agent is a low-code or even no-code AI platform. It integrates both structured and unstructured data and connects to various large language models (LLMs). This enables non-technical users to easily deploy GAI applications with simple drag-and-click actions, while IT personnel provide support by preparing data and assisting users with the WISE-AI Agent. This approach significantly accelerates GAI application development.

"The goal of WISE-AI Agent is to address the shortage of AI talent and enable manufacturers to leverage the low entry barriers of GAI for a faster path to digital transformation," stated Song. Traditional predictive AI, such as machine learning or deep learning, requires high implementation thresholds. Companies often need vast datasets to train AI models for accurate analysis, which deters many from adopting AI due to insufficient data. GAI, however, excels at reading and generating content from existing data, enabling rapid updates and feedback regardless of a factory’s automation level or dataset size, thereby significantly enhancing employee productivity.

Learn more about WISE-IoT solutions