A Neural Network Design Methodology for Structural Control
Abstract One of the major difficulties in neural network applications is the selection of the parameters in network configuration and the coefficients in learning rule for fast convergence as well as best system performance. This paper developed a network design methodology so that the optimal design parameters/coefficients can be determined in a systematic way thereby avoiding the lengthy trial-and-error. The methodology combines the Taguchi method of quality engineering and the back-propagation network with an adaptive learning rate for their advantages in implementation feasibility and performance robustness. Vibration suppression experiments of a composite smart structure with embedded piezoelectric sensor/actuator validate that the methodology provides an efficient neural controller design, including the plant order, the number of hidden layer neurons, the number of training patterns, and the coefficients of adaptive learning rate.