Digital Twin Modeling for Unmeasured Industrial Process States

The patent creates a digital twin (a virtual copy) of the physical industrial automation system to perform model training and verification. This copying approach allows the process model to be trained and verified in the virtual digital twin environment rather than requiring extensive computing resources on the actual physical system, thereby maintaining model accuracy while significantly improving computing resource efficiency.

The patent introduces a digital twin as an intermediary between the physical system and the control system. The digital twin acts as a mediator that receives process inputs and outputs from the physical system, performs model training and verification, and provides control recommendations back to the physical system. This intermediary approach enables determination of unmeasured operating states through the process model without directly complicating the physical control system architecture.

The patent implements periodic model training and verification cycles within the digital twin framework. Instead of continuous real-time model adjustments that would consume excessive computing resources, the system performs model training and verification at periodic intervals using historical process data accumulated during operation. This periodic action approach allows the system to optimize performance by updating the process model regularly while avoiding the continuous computing resource consumption that would result from constant real-time adjustments.