Structural Control-Deep Learning Applied to Acceleration Control of a Shaking Table

Structural Control-Deep Learning Applied to Acceleration Control of a Shaking Table

With the increasing complexity of equipment systems, determining their seismic resistance capacity simply through simulation is challenging. However, it is possible to directly test their capacity by a seismic shaking table. Due to the interactions between the shaking table and the specimen, controlling the shaking table becomes difficult, resulting in inaccurate reproduction of the history of acceleration response. To solve this problem, many scholars have used single-axis shaking table tests and have proposed new theories of control. However, significant opportunities for improvement of existing methods remain. Deep learning algorithms have the inherent capacity to model nonlinearity in dynamic systems. Therefore, the implementation of deep learning with respect to controlling a shaking table can improve the acceleration reproduction and thus are expected to facilitate development of new control methods.