In AI-powered digital twin applications, AI plays a key role at every layer and provides capabilities for:

– Advanced pre-processing and post-processing, hybrid simulation, data augmentation and fusion, and mining and discovery for the data layer.
– Hybrid AI modeling based on physics, data assimilation, downscaling, and post-model integration, among others, for the model layer.
– Recommendation, reasoning, and scenario generation under uncertainty, digital assistant, and visualization, among others, for the decision support layer.
The power of the DT+AI pairing lies in how the different AI support approaches can be combined to implement the DT layers. Without a fixed rule on how this integration should be done, developers can combine the approaches that best suit the particular needs to solve the problem at hand (in the particular DT layer) and connect them to compose a set of AI-compatible modules for a given configuration.