Neural Network Discrimination in Intelligent Vibration Control

The overall goal of this study lies in the multi-mode structural vibration control for systems, based on spatial information. In order to accomplish this task, a neural network was designed to identify the shape of the dominant vibration mode by using properly located piezoelectric sensors. The network then chose a set of scaling gains for a fuzzy logic controller that had been optimally tuned for that particular mode’s shape. Next, a fuzzy logic controller was employed to control the vibratory system. The novel technique was experimentally verified on an aluminum cantilever beam system that was augmented with two lead zirconate titanate (PZT) actuators and two PZT sensors. The optimized control strategy was then compared to traditional fuzzy logic control. Finally the mass of the plant was increased by over a factor of 1.5 and the same controller was used to adequately stop vibration.