A BP Neural Network Modeling Method Based on Global Error for the Hysteresis of Piezoelectric Actuator

5th Annual International Workshop on Materials Science and Engineering

Wang, Yanyan^1 ; Guo, Hai^2

Tianjin Key Laboratory of Information Sensing and Intelligent Control, Tianjin University of Technology and Education, Tianjin

300222, China^1

National Ocean Technology Center, Tianjin

300112, China^2

关键词: BP neural network model;    Connection weights;    Global errors;    Hysteresis curve;    Nano-positioning;    Positioning accuracy;    Training errors;    Training goals;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/585/1/012070/pdf
DOI  :  10.1088/1757-899X/585/1/012070

来源: IOP

【 摘 要 】

Piezoelectric actuator (PZT) is used widely in nano positioning, nano measurement and nano mechanics. However, its hysteresis, creep and nonlinearity affect the positioning accuracy seriously, especially the hysteresis. The paper proposes a BP neural network modeling method based on global error to model the hysteresis of the PZT. The network contains input, hidden and output layers. Its training goal is based on global errors. And the network could adjust the connection weight of the network dynamically according to different inputs till the global errors reduce to the threshold. Experiments prove that the method could fit the hysteresis curves of the PZT well. And the training errors could be controlled under 0.05.