Wavelet Network for Semi-Active Control

This paper proposes a wavelet neurocontroller capable of self-adaptation and self-organization for uncertain systems controlled with semi-active devices, ideal candidates for control of large-scale civil structures. A condition on the sliding surface for cantilever-like structures is defined. The issue of applicability of the control solution to largescale civil structures is made the central theme throughout the text, as this topic has not been extensively discussed in the literature. Stability and convergence of the proposed neurocontroller is assessed through various numerical simulations for harmonic, earthquake, and wind excitations. The simulations consist of semi-active dampers installed as a replacement to the current viscous damping system in an existing structure. The controller uses only localized measurements. Results show that the controller is stable for both active and semi-active control using limited measurements, and that it is capable of outperforming passive control strategies for earthquake and wind loads. In the case of wind load, the neurocontroller is found to also outperform an LQR controller designed using full knowledge of the states and system dynamics. Subject Headings: wavelet, neural networks, structural control, adaptive systems, control systems, intelligent structures, large structures, vibration mitigation