Machine learning is very promising in this area. With supervised learning you can create models from measurement data. This is especially attractive for complex nonlinear relationships. The resulting black box simulation models are often faster in execution and thus well suited for optimization.
Another application are virtual sensors. For example, you can use them to estimate forces or temperatures without using real sensors.
Reinforcement learning is very promising for optimising controls...
We have been using it successfully for active flow control (full text on ResearchGate and ArXiv also): https://www.cambridge.org/core/journals/journal-of-fluid-mechanics/article/artificial-neural-networks-trained-through-deep-reinforcement-learning-discover-control-strategies-for-active-flow-control/D5B80D809DFFD73760989A07F5E11039