Machine-Learning Based Optimal Active Seismic Control for Earthquake-Excited Building Structures

Machine-Learning Based Optimal Active Seismic Control for Earthquake-Excited Building Structures

In recent years, Linear-Quadratic Regulator (LQR) has been recognized as one of the most effective methods for structural vibration control which minimizes a cost function formulated by weighted states and control inputs. Optimal control requires structural states which may not be measurable in real application；therefore, state estimation is essential which inevitably takes additional computation time. However, time delay and state estimate error could affect the control performance. In this study, a multilayer perceptron (MLP) model and an autoregressive with exogenous inputs (ARX) model in machine learning are applied to learn the control force generated from an LQR with weighting matrices optimized by applying symbiotic organisms search algorithm. A shear building is adopted as a benchmark model for training and validation of the MLP and ARX models. In the numerical simulation, the dynamic analysis of the structure subjected to earthquakes is carried out with 8 sets of seismic accelerations to verify the performance of the controller. The results show that the neural network models are able to emulate the LQR control force from the acceleration response directly, reducing the necessity of a state estimator to achieve effective control performance in practice.