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.