Characterization of a piezoelectric-actuated, customizable-stiffness device for vibration suppression

Abstract

Vibration causes harmful effects for machines including total resonance failure, fatigue failure, user discomfort, and decreased performance. Because of these harmful effects creating effective vibration control measures is essential, especially for light weight structures used in aerospace applications. This research seeks to develop a customizable-stiffness device that could be usable in aerospace, vibration-control applications. The device is designed around a slender post-buckled beam with intermediate supports to influence its stiffness. Piezoelectric materials are used to actuate between two different support scenarios. There are three phases of this research a nonlinear, dynamic simulation; experimental testing; and a neural network analysis for design optimization. In the first phase, the primary concern was to obtain results for the post-buckling stiffness of the device. The first result was a nonlinear, dynamic simulation that provided consistent post-buckling results. Initially, a static simulation was expected to be sufficient, but it would not provide consistent results. From the dynamic simulation, results for the compressive force and displacement could be obtained to create a force-displacement diagram. By taking a numeric derivative, the stiffness results could be obtained. Secondly, these results suggested it was possible to affect the devices stiffness by changing the position of the supports. In the second phase, the primary concerns were proving physical efficacy of actuating the beam between two support scenarios and corroborating evidence for simulation results. Compression testing was completed in a custom buckling rig to obtain experimental force-displacement and stiffness-displacement curves. For the first two experiments, Macro Fiber Composites (MFCs) were bonded to the steel test specimen. The first experiment proved that an MFC was able to influence the direction of buckling. The second experiment demonstrated significant differences in the displacement-stiffness plot based on support positioning. The third experiment was useful in supporting the simulated results for loads at buckling and contacts. Additionally, the shapes of the simulated and real beam were compared. In the final phase, a neural network is created to predict results for the different support locations. A neural network was created to predict the stiffness of the device given a displacement and the location of the two supports. Using the stiffness predictions from the network, good force displacement diagrams. A root mean square error (RMSE) of 0.54 N was obtained from the best network. The stiffness-displacement diagrams did not align as well with a RMSE of 230 N/mm.