Micron Technology's two factories in Taiwan were recognized by the Global Lighthouse Network (GLN) under the World Economic Forum (WEF) in September 2020, according to the company's corporate vice president and Taiwan country manager Hsu Kuo-chin.
GLN is a group of advanced manufacturers promoting Industry 4.0.
While the development of 5G and AI leads to demand for large volumes of data, memory chips are becoming increasingly complicated, with better performance, less power consumption and small dimensions, Hsu said, adding memory chip makers have to hike precision of products and differentiate manufacturing processes by virtue of smart manufacturing technologies.
Micron began to undertake smart manufacturing at its factories globally in 2015, Hsu said. Initially, data were internally integrated under an architecture enabling visualization and transparency; then in 2017 various IoT technologies were put into use for real-time monitoring, Hsu indicated, adding that engineers now can predict abnormal conditions using AI.
The smart manufacturing began at wafer fabrication and has been extended to IC packaging and testing, procurement and supply chain manufacturing, Hsu said. In particular, wafer fabrication involves more than 1,000 manufacturing processes and takes three months, while IC packaging and testing take only 3-5 days, Hsu noted.
Currently, Micron collects data of 13TB in total from all factories a day, equivalent to 6,500 hours of movies, and uses AI to analyze 3.3 million images of wafers a week, Hsu indicated. In order for real-time monitoring, Micron has installed a total of 110,000 sensors at its 13 factories around the world, Hsu said.
The smart manufacturing is intended to boost productivity through using AI to simulate scheduling and optimize manufacturing processes, with AR/VR technologies to support remote maintenance of equipment and robots. Detection of operating indexes and anomalies at production lines is enabled by real-time monitoring based on IoT, machine vision and deep learning technologies. Automatic recognition of and response to anomalies through using an in-house-developed AI-based analytical and diagnostic system can hike yield rates, Hsu explained.
For manufacturing processes in particular, Micron uses sensing technologies to detect dynamic and static images as well as sounds to real-time monitor operating conditions of equipment and product quality, Hsu said.
The smart manufacturing has resulted in 18% increase in employee's productivity, 34% decrease in unplanned stoppage of equipment in operation, and 40% reduction in non-conforming products, Hsu noted. In addition, time taken in hiking yield rates for new products has been reduced by 20% and total power consumption by 15%, Hsu added.