Intelligent digital triplets for autonomous machine optimizations

Abstract

This project presents a novel technology named Intelligent Digital Triplet (IDTR) to overcome the traditional Intelligent Digital Twin (IDT) issues. These issues are: reconfigurability, mission critical services, portability, and automation of time critical processes. IDTR targets complex manufacturing processes, and it improves their overall productivity. Also, this project aimed to design a new Digital Triplet system within a semiconductor manufacturing facility. While an Intelligent Digital Twin (IDT) is a virtual replica of a process, product, or service with an Artificial Intelligence (AI) module, the Intelligent Digital Triplet (IDTR) builds upon an IDT and goes beyond its capabilities. The novelty of this project is in providing a fully autonomous virtual environment to act as a triplet replica to improve the performance of traditional IDT systems. More specifically, adding a feature to this new Digital Triplet system so that it becomes a self-trained virtual environment that learns from the errors it identifies and takes actions to rectify them. Consider a manufacturing process that is not producing the desired yield. It receives the same inputs which the system obtains and displays the same output as the live system. Traditional IDT systems would analyse the data and highlight the problem. With some more data crunching, it might also propose some changes to the system. The Intelligent Digital Triplet (IDTR) builds upon an IDT and goes beyond its capabilities. It loads a virtual replica of the live process inside a sandbox that is referred to as a triplet in the developed system. By using AI techniques, it starts optimising the virtual system and then analyses the effects of the changes. It only proposes changes once the AI model has tried the new parameters, and it is confident of their effectiveness. The AI algorithms used by the IDTR are intelligent agents that virtually roam the IDTR environment, in a manner similar to Hummingbirds, and we named them Intelligent Digital Humming Tools (IDHT) in this project. They are AI agents that are capable of taking autonomous decisions, reacting to the environment’s stimuli, acting proactively, and communicating with each other. IDHT tools roam in the virtual environment to observe pre-determined conditions and parameters about the different elements of the virtual production environment. While the IDT visualizes real-time insights about the process to explain what happens in the physical production lines, the developed IDTR system tackles any predicted behaviours and takes more effective autonomous decisions. This project uses a case study of an international semiconductor manufacturing company (STMicroelectronics) to digitize and automate their physical production processes. The developed IDTR facilitates product design, improves the planning and incorporates more effective predictive maintenance practices. The developed IDTR system showed great potential in root-cause analyses and predictive maintenance applications. IDTR improved the overall prediction accuracy of the IDT from 78% to 93.8%. It has also improved the accuracy of Root Cause Analyses capabilities from 88% to 94.2%. Also, the developed IDTR system reduced unplanned equipment’s downtime by 40%. Overall, such improvements could potentially lead to saving the adopters millions of Euros.