Research On Artificial Neural Network Application In Active Damping System Using Piezoelectric Actuators

Wind tunnel models are usually mounted on long cantilever stings in order to reduce supportsystem flow interference on data. Sting vibration is caused by aerodynamic load changes. Such stingoscillations are known to affect data accuracy in many test situations. To eliminate the vibration, anactive damping system based on neural network PID control method and piezoelectric actuators wasdeveloped and evaluated through experiment. This paper presents an exploratory research oncantilever sting vibration control, the contents are as follows:(1) By simplifying the cantilever sting system and according to material parameter, modal analysiswas conducted and the first six modes of the sting system were obtained.(2) Through processing and analyzing the balance signal of wind tunnel test, the main vibrationmode and the aerodynamic range as well as actuators position were determined. Then the force on theactuators was calculated and the sting was redesigned to fix them. Modal analysis of new stingstructure with piezoelectric actuators was conducted again.(3)A neural network PID control method was applied which can adjust control parametersautomatically. Besides the controller based on LabVIEW software platform and NI PCIe7841Ronboard data acquisition card was developed, the cantilever beam vibration damping system was thenset up to validate and optimize the control algorithm.(4) Then the performance of the damping system based on neural network PID controller wasevaluated through both ground and wind tunnel experiment. It was shown that, respectively forground and wind tunnel experiment, the decay time was shortened by50%and45%, vibrationamplitude was damped by56%å'Œ45%respectively compared to conventional PID, and the first modeamplitude was damped at least19dB under various wind velocities and attack angels.