The Modeling And Control Of Piezoelectric Workbench Baseed On Neural Network

In recent years, with the development of industrial fields, such as micro machinerymanufacturing, precision measurement, ultra precision processing, integrated circuitmanufacturing, biological engineering, medical science and other fields, The demand ofthe related equipment in the location of stroke and the positioning precision becomesmore higher, so the micro positioning technology has been one of the hot spots in theprecision engineering field. Micro positioner drivd by piezoelectric element is thecommon micro positioning device, and the piezoelectric workbench is one typicalrepresentative.Piezoelectric workbench can use piezoelectric element piezoelectric orelectrostrictive effect to achieve micron even nanometer precision positioning. Most ofthe piezoelectric bench have the merit of micro displacement of high resolution, strongstability, fast response, high rigidity and can be miniaturized and other advantages,because of the piezoelectric ceramic stack and a flexible hinge support of combinedtype mechanical design structure that appiying inside, and as its hardware sex, it canachieve a perfect physical connection with all kinds of objections, such as sensors,computer, Power amplifier and etc in closed loop control., and adding its volume whichis smaller than any of the other micro displacement locator., it has been widely appliedin various fields. However, the hysteresis, creep, nonlinear of the piezoelectricworkbench caused by some properties defects of ceramic materials, reduce the locationaccuracy and the dynamic response speed，which results a certain displacement error. Inorder to obtain the requirements of the location accuracy and the dynamic responsespeed. It’s required to research the characteristics defects of piezoelectric ceramic anddesign some corresponding control model and controller.According to the double sigmoid activation function though of XiangDong LiuProfessor.,the paper adapts and packages some corresponding functions in MATLABneural network, and applies the adapted neural network function to model, it researchesthe piezoelectric offline model and piezoelectric online model separately, and thendesigns an adaptive neural network controller to improve the positioning precision of Piezopositioning stage according to the research of the model. At the end of the paper,it tracks the triangular wave and complex frequency wave for the designed model andthe controller to verify the model precision together with control precision. And theexperimental effect is satisfied.