Machine Learning for Hydraulics Components Capley Shaver

We ask ourselves the question at the very beginning whether it's possible to apply machine learning to an hydraulic system. The RL agent tried to predict new values for our PI D controller and so the PID controller was able to adapt to changing working points throughout black the whole operating process. We never trained the reinforcement learning agent on a real test bench because we've seen in the simulation that it would require data from more than 24 hours of operating time. This doesn't make sense energy wise but it also wouldn't be good for the hardware as like when you have reinforcement learning you have the exploration phase where the algorithm explores also like system boundaries and all that stuff.