Focus on Industrial AI: Ming Chi University’s Outstanding Achievements
With the rapid evolution of AI algorithms and advancements in semiconductor processing, AI has become a top choice for companies aiming to strengthen their competitiveness. Formosa Plastics Group, which spans industries such as petrochemicals, power generation, electronics, silicon wafers, biotechnology, healthcare, and steel, declared 2019 as its “Year of AI,” integrating AI into its production processes through various channels. According to data released by the Formosa Plastics Group, AI has been applied in areas such as intelligent manufacturing, quality inspection, and process optimization, completing 318 projects by the end of 2020 with an estimated annual benefit of NT$2.134 billion. Currently, 336 projects are ongoing, projected to yield an additional NT$1.824 billion annually upon completion.
Formosa Plastics Group has demonstrated impressive results with AI technology, with each business unit establishing AI research centers to promote AI applications. In 2019, the group’s Ming Chi University of Technology established the Center for Artificial Intelligence & Data Science (AI & DS Center) to support various AI projects within the Formosa Plastics Group. The center is dedicated to advancing AI technology, cultivating AI talent, and conducting industry-academia collaborations with various business units within the Formosa Plastics Group, positioning itself as a vital contributor to AI application development.
Ching-Shih Tsou, Director of Ming Chi University of Technology’s AI & DS Center, explained that the center was established to support the Formosa Plastics Group in applying data science (DS) technologies to solve challenges that various business units face in developing AI, thereby enhancing overall competitiveness. Supported by various resources, the AI & DS Center was established as a research center. In addition to collaborating with Formosa Plastics Group through industry-academia partnerships, the AI & DS Center also assists other enterprises with data analysis and modeling needs. From an educational standpoint, the center is committed to cultivating AI talent and providing AI knowledge education to nurture the industry’s AI talent.
Cross-Domain Integration: A Synergy of Software and Hardware, Supported by Data Science
Yen-Chen Chen, Deputy Director of Ming Chi University of Technology’s AI & DS Center, mentioned that experienced technicians can often identify potential faults in equipment simply by listening to its operational sounds. However, this knowledge is not easily passed down. By recording equipment operating sounds and applying AI technology for analysis, combined with technician insights, the center has designed a comprehensive predictive maintenance mechanism.
In response to the rapid development and widespread adoption of AI technology, Ming Chi University established the AI & DS Center in August 2019, with over 103 pings (approximately 341 square meters) of space and nearly NT$50 million in funding. This center, equipped with an NVIDIA DGX-1 AI supercomputer, meets the needs for office and research space. The center focuses on industrial AI research, with an emphasis on combining domain expertise and AI technology to address practical industry challenges through industry-academia partnerships and to cultivate AI+X talent (Artificial Intelligence + domain Expert). Since its establishment, the AI & DS Center has undertaken over NT$50 million in AI-related industry-academia projects, with many more projects in progress, achieving remarkable results.
In terms of industrial AI technology, there are six major areas: “failure prediction and preventive maintenance mechanism,” “optimization of material combinations, scheduling, and manufacturing processes,” “design and integration of private cloud, sensors, and IoT,” “industrial image recognition and quality inspection system architecture,” “business intelligence and management enhancement mechanism exploration,” and “safety inspection and accident prevention.” In terms of talent cultivation, the school offers foundational AI courses and specialized credits or degree programs, along with knowledge-oriented and project-oriented training courses tailored to industry needs to assist in training the required workforce.
Yen-Chen Chen emphasized that equipment maintenance has always been an important task in factories, and experienced technicians may recognize potential failures just by listening to equipment operation. However, this experience is challenging to standardize and quickly pass on to new employees. By recording the operational sounds of equipment and using AI to analyze them, combined with technician knowledge, the center has developed a robust predictive maintenance mechanism. This helps avoid issues such as shutdowns, safety incidents, or wastage of raw materials or semi-finished products due to equipment failures.
Additionally, chemical plant production is highly automated, and variations in material combinations or timing of raw material inputs can significantly impact product quality. AI can help identify the best production combinations and automate procedures, greatly enhancing production efficiency. With its significant production in the chemical industry, Formosa Plastics Group naturally seeks assistance from the AI & DS Center to improve overall competitiveness in the chemical sector.
Training AI+X Talent to Meet Diverse Industry Needs
Ching-Shih Tsou stated that the AI & DS Center primarily focuses on industrial AI research, and its team is divided into four major groups. The Unit Equipment Group focuses on the design, integration, and arrangement of industrial sensors and the integration of mechatronic control systems for subsequent defect signal extraction, separation, and intelligent diagnostic analysis. The Process Systems Group specializes in optimizing parameters in processes such as chemical and plasma processing, process simulation environments, process safety and application evaluation, and planning and layout of pilot production environments.
The Data Analysis Group focuses on big data analysis across various data types, integrating data cleaning, feature extraction, time series, machine learning, and deep learning algorithms to analyze data such as production logs, industrial signals, industrial images, surveillance footage, document text, network relationships, and positioning records. Meanwhile, the Program Development Group focuses on nurturing AI+X talent, infusing AI knowledge into courses across various departments and creating customized training courses based on industry needs.
Ching-Shih Tsou pointed out that training AI+X talent means cultivating AI experts who also possess expertise in their specific application fields. For instance, Formosa Plastics Group operates a power plant in Mailiao, Yunlin. Developing a predictive maintenance mechanism based on motor operation sounds would require assistance from mechanical experts. The goal of AI talent cultivation at the AI & DS Center is to apply AI perception along with data analysis models from science, engineering, and business management to create interdisciplinary AI+X project teams that address industry challenges through collaborative projects. This provides students with industry-relevant experience and reduces the gap between academic learning and practical application.
A Balanced Approach Combining Theory and Practice
A synthetic fiber plant frequently encounters quality inconsistencies due to variations in raw material batches and interference during batch changes. They sought assistance from the AI & DS Center, which applied its AI expertise, equipment, and expertise in chemical engineering to develop a predictive quality model, achieving consistent quality control across products.
Ching-Shih Tsou mentioned that the AI & DS Center collaborates with various departments within the university and is in the process of establishing a bachelor’s degree program in Industrial Artificial Intelligence (IAI). This program combines faculty from different fields with industry experts, forming an AI+X dual-teacher team that focuses on interdisciplinary courses, diverse faculty, and industry internships. Real-world projects with partner companies serve as training grounds for students both on-campus and in off-campus internships.
9 AI Servers and 60 AI Workstations
With growing research results and an expanding scope of industry-academia collaborations, the scale of Ming Chi University’s AI & DS Center continues to expand. The center currently has three NVIDIA DGX-1 AI computers and six DELL servers, totaling nine servers with a combined 42 NVIDIA V100 (or A100) GPUs, providing a high-stability platform for deep learning and data processing. By using Infinitix’s CloudFusion hybrid cloud management software, virtualized resources can be dynamically allocated to researchers, allowing GPU-trained models to accelerate project testing and research innovation.
The center also has a virtualized classroom environment with about 40 pings (approximately 133 square meters) for AI teaching, equipped with servers, terminal hosts, virtualization machines, as well as broadcasting and control management functions. This educational environment, built using x86 servers, Ubuntu, and VirtualBox, is primarily used for machine learning and data processing platforms. In addition to projects and research, it allows up to 60 students to experience high-speed GPU training.
Yen-Chen Chen explained, “The center primarily uses powerful open-source software such as Python, R, and C# to develop project code and user interfaces. For non-programming specialists, we have commercial software suites such as LabView, OpenR8, and STATISTICA 13, encouraging introductory learning and reducing programming barriers.” Additionally, the center has established a 23-ping (approximately 76 square meters) AI Situation Room, where the team can use the CRESTRON CP2E flexible control system to quickly integrate situation room hardware and software for AI demonstration, utilizing wireless control panels and presentation controllers for smooth presentations.
Introducing AI-Stack Platform to Maximize GPU Resources
Reflecting on the initial establishment of the AI&DS Center at Ming Chi University of Technology, while high-performance NVIDIA DGX-1 AI computers were procured, the system maintenance engineers, facing heavy workloads, were not always familiar with all open-source tools or software packages used by project personnel. This led to considerable time being spent on resource allocation. To address this, the AI-Stack by Infinitix, a platform that effectively manages DGX-1 resources and enables flexible scheduling, was adopted. Through this platform, NVIDIA DGX-1 AI computers are managed, allowing GPU computing resources to offer flexibility, high-efficiency collaboration, and improved operational cost-effectiveness.
According to Cheng-Shih Tzou, as the AI&DS Center continues to expand, the number of NVIDIA DGX-1 AI computers has increased from one to three. AI-Stack has successfully connected these three AI servers into a controllable, manageable, shareable, and horizontally scalable machine learning/deep learning computational resource pool. Based on this foundation, students, teachers, researchers, and IT administrators alike can rely on a user-friendly interface, a straightforward AI learning environment, and convenient allocation and monitoring capabilities to facilitate AI research and teaching while maximizing valuable GPU resource utilization.
As AI technology continues to drive digital transformation and enable various intelligent services worldwide, there is a growing talent shortage in the AI field. Consequently, the AI&DS Center at Ming Chi University of Technology has plans to extend its support beyond Formosa Plastics Group in the future. This includes offering AI and business model integration services, establishing dedicated AI education programs, and fully supporting businesses in their use of AI applications.
