Noncolocated vibration control using neural networks

Abstract

We have previously successfully developed collocated neurocontrollers for vibration suppression. The objective of this research is to develop neural network controllers for vibration suppression under the constraint of noncollocated control. Noncollocated control may be necessary under at least two scenarios: physical constraints of sensor/actuator placement, or reconfiguration due to damaged elements. The neural controller is based on forward dynamics and decentralized control. This controller has been demonstrated to be effective in vibration suppression for two different systems.