Development of a Control Algorithm for a Semi-Active Mid-Story Isolation System Using Reinforcement Learning

The semi-active control system is widely used to reduce the seismic response of building structures. Its performance mainly depends on applied algorithms. Various algorithms have been developed date. Recently, machine learning has various engineering fields and provided successful results. Because reinforcement (RL) shown good for real-time decision-making problems, structural engineers become interested in RL. In this study, RL was development a algorithm. Among methods, Deep Q-network (DQN) selected because its application many problems. A sample structure constructed by using mid-story isolation (SMIS) with magnetorheological damper. Artificial ground motions were generated numerical simulation. excitation make environment. reward designed peak story drift drift. Skyhook groundhook comparative study. Based results, paper shows that proposed algorithm can effectively responses structures SMIS.