STMicroelectronics has built a digital twin of its fab in Crolles, France, to monitor and optimise equipment.
The company is using its STM32 32bit microcontrollers with embedded AI to monitor fab equipment and automated transport systems. This is combined with multiple data sources in a digital twin that allows predictive maintenance and system optimisation, says Philippe Magarshack, Group Vice President of the Group Vice President of Strategic Technology & System Architecture in the Microcontrollers and Digital ICs Group (MDG) group at ST. He is also vice-president of the Minalogic Collaborative R&D Cluster in Grenoble.
“Industry 4.0 applications are very diverse and require sophisticated CPS hardware and software, from security, cloud, security, wireless and RTOS,” said Magarshack in a keynote on cyberphysical systems (CPS) at the DATE 2021 conference. “In smart factories all production steps are digitally mapped in a digital twin and connected. At ST we are not only designing products for Industry 4 .0 but using these techniques. Several hundreds of steps in a fab with multiple parameters so it is fundamental to collect and analyse in real time all these parameters.”
Running 6300 wafers per week the 300mm fab at Crolles generates 3TB of data per day. ST is using three AI techniques for monitoring wafers and equipment. Advanced defect classification (ADC) automatically classifies wafer defects with CNN neural network pattern matching so that corrective action in real time and defects are caught before other processes are added in. Then there is equipment fault detection. Instead of measuring a single temperature the complete curve is captured and compared against the golden curve. This allows real time modelling and decision making to optimise productivity, says Magarshack.