The industrial world is in constant evolution, with new technologies playing an essential role in this transformation. One of the major advancements is the emergence of the hybrid digital twin, an innovation promising to revolutionize the realm of industrial maintenance.
A digital twin is a computerized replica, digitally modeled to be in sync with a real machine or process. The digital twin can be utilized to predict maintenance needs, detect faults, and anticipate the consequences of such events. Traditionally, it relied on machine learning, but today, the "hybrid" approach aims to merge machine learning with physical simulation models. "The potential of the hybrid digital twin lies in combining these two approaches," explains Beat Wolf, an assistant professor at the University of Engineering and Architecture (HEIA-FR) and an expert in machine learning applied to industrial products.
The hybrid digital twin aims to transition from more general predictive maintenance to more targeted and efficient prescriptive maintenance. Predictive maintenance focuses on precursor signals of failure without necessarily identifying the source. On the other hand, prescriptive maintenance aims to identify the sources of dysfunction and propose corrective actions to prevent them. "Thanks to the physical model, the hybrid digital twin provides a deeper understanding and additional information about the nature of detected faults," emphasizes Jean-Luc Robyr, an associate professor at HEIA-FR and an expert in physical signal analysis.
Another strength of the hybrid digital twin lies in its autonomy. Running in parallel with the real machine, it can detect abnormal behaviors and characterize the machine's condition. It can foresee failures and take measures to prevent or even correct them on its own, thereby reducing human involvement in maintenance. "The hybrid digital twin can be seen as an ever-vigilant expert, constantly analyzing machine signals," explains Beat Wolf. He illustrates the potential of the digital twin's automation with the example of a printer: "When the print head of a printer shows signs of blockage, the digital twin can anticipate this issue and automatically trigger a cleaning cycle to avoid the problem. Instead of performing periodic recalibrations and cleanings, even when not necessary, the machine can now autonomously carry out these operations at the opportune moment."
To fully explore the potential of the hybrid digital twin, a collaborative project named ModIA has been launched. The project brings together HEIA-FR and several Fribourg-based companies with various industrial machines. The main objective of ModIA is to practically implement a prescriptive maintenance approach and study its specific requirements. Three companies involved in the ModIA collaborative project will develop their own hybrid digital twin tailored to their machine types, while the other two companies will contribute their technological expertise.
By combining physics and artificial intelligence, the hybrid digital twin aims to achieve more efficient prescriptive maintenance. Through its autonomy and automation potential, it relieves humans from repetitive and time-consuming tasks. "This provides us with a better understanding of the machine's condition and allows us to propose specific actions to correct detected issues, significantly enhancing maintenance efficiency," explains Jean-Luc Robyr.
The ModIA collaborative project is a concrete example of this technology in action, creating collective value and paving the way for a deeper understanding and optimization of industrial systems.