Neural Network for Structure Control

Abstract

Significant progress has been achieved in the active control of civil-engineering structures, not only in the control algorithm, but also in the control testing of the scale model and full-scale building. At the present time, most algorithms used in the active control of civil-engineering structures are based on the optimization of the instantaneous objective function. In this paper, a Backpropagation-Through-Time Neural Controller (BTTNC) developed for active control of structures under dynamic loadings is presented. The BTTNC consists of two components: (1) a Neural Emulator Network to represent the structure to be controlled; and (2) a Neural Action Network to determine the control action on the structure. The artificial neural-network controller is a newly developed technique for the purposes of control and has many attributes, such as massive parallelism, adaptability, robustness, and the inherent capability to handle nonlinear systems. Results from computer-simulation studies have shown great promise for the control of civil-engineering structures under dynamic loadings using the artificial neural-network controller.