Workshop

"Pre-operation models (mathematical and/or 3D) are often used as Digital Twins of new things to be built. They allow testing and experimenting various configurations and operational parameters before the elements are built, assembled, and put in operation. Assessments can then be made on the suitability of the design placed into its context. This method, however, only allows to make the best theoretical assumptions. As close to the reality as they can be, the system in operation will always behave slightly differently. Indeed, systems are composed of parts, that follow slightly different life cycles. Repairs and maintenances can happen in some cases and not in others. Said otherwise, two initially identical pieces of equipment will have slightly different behaviors with time.

DataThings' approach uses data from past experiences to create digital twins of systems in production. With this, these digital twins account for the specificity of the systems and their context of operations, to deliver simulations and predicted responses closer to the reality. The digital twins created and learned from operational data can then be used to better anticipate the actual systems' reactions and support the day-to-day operational decision making. However, those twins can only provide meaningful predictions when faced with parameters within the range of the operations history on which the learning has been performed. Exploring the boundaries can only be achieved with real-life experiences.

In this workshop, we will start with presenting the general approach used to learn from past experiences, and the different learning mechanisms that can be used for different purposes. In a second part, we will provide more details on the implementation of this approach for WP2 Process; and for WP3 Product in a third part."

Check out the video of the workshop: Youtube