-
Lesson 6: Deep Reinforcement Learning (DRL)
Some of my queries are listed below:
1. What is the use of deep reinforcement learning in multiphase flow? Can it help improve hydrodynamic performance or predictions? I am not clear about it. Could you explain the concept using any multiphase flow system as an example?
2. Firstly, I am not able to understand the exact meaning of the “state” in reinforcement learning
3. What is the meaning of control strategy, and why is it used?
4. In multi-agent reinforcement learning:
4.1. What is the physical meaning of state (u,v)?
4.2. Why is the reward expressed in terms of stress, while the action is expressed in terms of velocity at the wall? And how were these states, actions, and rewards selected?
4.3. Are all agents similar in multi-agent reinforcement learning?
5. Could you also refer to some hands-on material to practice DRL?
6. Reducing the Nusselt number using the multi-agent and single-agent DRL would also be validated through experiments.
<b style=”cursor: text;”>7. All the examples show single-phase flow and a small-scale system. Could this approach be feasible in large multiphase flow systems?
<b style=”cursor: text;”>8. Would DRL be a computationally time-efficient approach?
9. Would DRL work with Eulerian-based CFD simulations, given that it does not include microscale analysis?
-
This discussion was modified 2 months ago by
Manoj Kumar Beriya.
-
This discussion was modified 2 months ago by
Sorry, there were no replies found.
Log in to reply.
